Impact & Recognition

Federal Register May 2026

Cited in the U.S. Federal Register (NHTSA)

National Highway Traffic Safety Administration · 91 FR 30789 · Document 2026-10363

A public comment by the applicant (Docket NHTSA-2026-0529-0007, Apr. 27, 2026) is cited by name in an official U.S. Federal Register notice: NHTSA, "Incident Reporting for Automated Driving Systems (ADS) and Level 2 Advanced Driver Assistance Systems (ADAS)," 91 FR 30789 (May 26, 2026). The notice references public comments received on its automated-vehicle incident-reporting collection; among the submissions referenced are those of Virginia DOT, the Alliance for Automotive Innovation, and Consumer Reports. The citation is permanent and independently verifiable through the Office of the Federal Register and govinfo.gov.

Source: NHTSA, 91 FR 30789 (May 26, 2026), Document 2026-10363. Verify on federalregister.gov · Download PDF (the official notice; his comment is cited in footnote 10, on page 30792)

arXiv Preprint May 2026

Cited in an arXiv preprint on shadow AI in critical infrastructure

Baruwal Chhetri, Tariq, Aamir, Grobler, Thapa & Singh · arXiv:2606.00088 · Submitted 23 May 2026

"From Frontier to Shadow AI: A Simmering Threat to Assurance and Security in Critical Infrastructure," an empirical study of 27 Australian critical-infrastructure organisations, cites the applicant's SSRN paper The strategic convergence: AI has outpaced human clearance models (SSRN 5940814) as reference [41]. The applicant's work is cited twice in the body: on how shadow AI operates at the interaction layer, distinct from traditional shadow IT, and on shadow AI as an insider-risk condition driven by optimisation-oriented workarounds. View on arXiv.

arXiv Preprint July 2026

Cited in an arXiv preprint on unsupervised tunnel detection in ground-penetrating radar

Junaid, Khan & Ahmed · arXiv:2607.04882 · Submitted 6 July 2026

"Unsupervised Detection of Underground Tunnels in Ground-Penetrating Radar Using Depth-Restricted Reconstruction Scoring," a field study by researchers at Sir Syed CASE Institute of Technology (Islamabad, Pakistan) and the University of Strathclyde (Glasgow, UK), cites the applicant's SSRN paper Strategic Subterranean Domain Awareness: A Comprehensive Technical and Operational Evaluation of Next-Generation AI-Fused Counter-Tunnel Architectures (SSRN 6046594) as reference [1]. The applicant's work is cited twice in the body: on the limits of conventional pipeline monitoring as early warning against clandestine tunnel excavation, and on why intrusion tunnels pose a different detection problem from buried pipes and cables in scale, depth, and rarity. View on arXiv.

Federal Technical Evaluation June 2026

White paper favorably evaluated by the Air Force Research Laboratory under a competitive Broad Agency Announcement

AFRL/RFC, Rome Research Site · Official memorandum of June 24, 2026 · BAA FA8750-24-S-7003 (CANVAS)

The Air Force Research Laboratory’s Information and Spectrum Warfare Directorate issued a formal memorandum responding to the applicant’s white paper, “Conditions-Based Authority Governance and Assurance for the Virtual C2 Layer: Verified Delegation, Authority Tradeoff Analysis, and Intent-Constraint Validation,” submitted under Broad Agency Announcement FA8750-24-S-7003 (CANVAS: Coordinating Austere Nodes through Virtualization and Analysis of Streams). The memorandum, signed by the Chief of the Multi-Domain C3 Division, states that the white paper “has been favorably evaluated” and that the effort described “is of interest to the Air Force,” while declining to fund it at this time due to budget constraints, noting that a formal proposal may be requested if the funding situation changes, and encouraging submission of additional white papers in areas of demonstrated capability. The memorandum documents an arms-length technical evaluation by a federal research laboratory of the applicant’s authority-governance research as applied to the virtual command-and-control layer. It is recorded here as an official evaluation outcome; it does not constitute AFRL endorsement of, or affiliation with, the research program, and no funding was awarded.

Conference Presentation September 2026

Presented at the NDIA Emerging Technologies for Defense Conference and Exhibition

National Defense Industrial Association (NDIA) and its Emerging Technologies Institute · Washington, DC · Presented September 2026

The applicant presented his runtime authority-governance research at the National Defense Industrial Association’s Emerging Technologies for Defense Conference and Exhibition in Washington, DC, in September 2026, an event convened with the association’s Emerging Technologies Institute. Acceptance to present at a national defense-industry technical conference proceeds through a selective programme-committee process and places the work before defence programme offices, prime contractors and technology evaluators. The question he presented is the one his architectures are built around and the half of the assurance problem that receives least attention: when an autonomous system loses authority because confidence has degraded, WHAT EVIDENCE SHOULD BE REQUIRED BEFORE THAT AUTHORITY IS RESTORED. His position, published across his journal and commentary record, is that reducing authority is only half of the problem, and that restoration should be governed by explicit and testable evidence rather than by the return of a single apparently clean signal. He presented in the conference’s digital poster programme, with the display shown on the main exhibit floor throughout the day, and he has recorded that the discussions around it were useful in identifying where the concepts meet real operational constraints, which is the engagement a conference of this kind exists to produce.

International Engagement July 2026

Verification working papers circulated to CCW state delegations; invited consultation with the Netherlands Ministry of Foreign Affairs

United Nations Convention on Certain Conventional Weapons (CCW) framework · July 2026

The applicant’s verification and compliance-attestation working papers for autonomous systems (Zenodo DOIs 10.5281/zenodo.21307610 and 10.5281/zenodo.21307785) were circulated to state delegations within the CCW framework ahead of the Seventh Review Conference, and the applicant received an invited technical consultation with the Netherlands Ministry of Foreign Affairs team supporting the Chair of the Group of Governmental Experts (July 2026). The papers address verification and compliance attestation, consistent with the program’s governance-and-assurance-only scope; circulation and consultation are documented engagement and do not constitute adoption of, or governmental endorsement of, the applicant’s work.

Working paper I ↗ Working paper II ↗

Institutional Response 2026

Written technical responses received from the RAND Corporation and NASA

RAND Corporation; National Aeronautics and Space Administration · Correspondence on file, 2026

In response to the applicant’s technical outreach, researchers at the RAND Corporation and at NASA provided substantive written technical responses engaging with the runtime authority-governance research; the RAND response confirmed the work’s relevance to the recipient’s research area and noted internal distribution to colleagues. The correspondence is retained on file and is available for verification in appropriate contexts. These responses document professional engagement with the work by researchers at the named organizations and do not constitute endorsement by, or affiliation with, either organization.

Peer-Reviewed Journal Article July 2026

Peer-reviewed article published in the Journal of Hardware and Systems Security (Springer)

Journal of Hardware and Systems Security (Springer Nature) · Vol. 10, Art. 17 · Received 4 June 2026, accepted 14 July 2026, published 21 July 2026 · Sole author: Burak Oktenli

Springer’s Journal of Hardware and Systems Security published the applicant’s peer-reviewed article “SATA: Sensor Attestation and Trust Anchoring: A Formally Specified, Simulation-Evaluated Continuous-Trust Protocol for Autonomous-Systems Sensor Edges” (DOI 10.1007/s41635-026-00190-4). The article specifies a TPM 2.0 hardware-anchored protocol that computes a continuous trust scalar from a sliding window of cryptographically committed sensor-attestation records, combining six replay-resistance barriers (ECDSA P-256 signature, 256-bit CSPRNG nonce, TPM monotonic counter, PCR quote, per-sensor sequence, and hardware tick-clock age) with a stated-and-proved signature-barrier unforgeability bound under a formally defined A1/A2 adversary model, and a weighted Dempster-Shafer fusion layer providing Byzantine fault tolerance while preserving the ignorance-versus-distrust distinction. The protocol’s state machine is verified by TLA+ bounded model checking across 18,892 distinct reachable states with zero violations of eight safety invariants, complemented by a 10,000-run Monte Carlo conformance-and-sensitivity campaign with seeded, configuration-exact reproducibility; the reproducibility artifacts are openly deposited on Zenodo (DOI 10.5281/zenodo.18936251) and are content-addressable through published cryptographic hashes. This is the peer-reviewed instantiation of the SATA architecture within the applicant’s runtime authority-governance research program, and it constitutes independent scholarly validation of that work through formal peer review.

Read at Springer ↗ Reproducibility artifacts (Zenodo) ↗

Scholarly Dissemination As of July 21, 2026

SSRN dissemination metrics: 17 public papers, 8,395 downloads, author rank in approximately the top 0.5 percent

SSRN (Elsevier) · Author ID 9424552 · Snapshot of September 11, 2026

The applicant’s SSRN author record lists 17 publicly distributed working papers spanning authority governance for autonomous systems, AI-enabled command and control, critical-infrastructure resilience, space and quantum-communications governance, and counter-tunnel and undersea detection analysis. As of September 11, 2026 the papers have accumulated 8,395 downloads; abstract views stood at 5,604 in aggregate at the July 21, 2026 snapshot, with the most-viewed paper, “Strategic Assessment: The Operationalization of Artificial Intelligence in U.S. Defense Doctrine,” exceeding 1,200 views and 740 downloads. SSRN ranks the applicant 14,540th of 2,819,733 registered authors, placing him within approximately the top 0.5 percent of authors on the platform by SSRN’s ranking methodology. These figures are point-in-time values from SSRN’s public author record and change as the platform updates.

SSRN author page ↗

Independent Citation August 2026

Translated into Albanian and republished by a Balkan geopolitics institute

Instituti Shqiptar për Gjeopolitikë (Albanian Institute for Geopolitics) · Pristina · Analysis section · August 22, 2026

The Albanian Institute for Geopolitics, a non-governmental research institution founded in 2021 and based in Pristina, translated the applicant’s Eurasia Review analysis of drone-war testing standards into Albanian and republished it in full in its Analysis section, crediting the original outlet in the byline, linking his author page there, and marking the translation as the institute’s own work. The institute states that it conducts interdisciplinary research in geopolitical and historical studies and European and Euro-Atlantic integration, and that its membership comprises scientists, publicists, and journalists drawn from Kosovo, North Macedonia, Albania, Slovenia, Switzerland, Germany, Austria, and the United States. The applicant’s analysis appears in the same section as the institute’s coverage of regional security questions and alliance posture. This is the first recorded instance of his work being translated into another language and republished in full by a research institution, and it is a distinct form of third-party use from syndication or excerpting, since translation requires an editorial judgement that the argument is worth carrying to a readership that would not otherwise reach it, and the institute performed and credited that work itself. His full author biography was carried in translation alongside the article.

Read the Albanian edition ↗

Accepted for Publication September 2026

Manuscript on governance engineering accepted by the Naval War College Review

Naval War College Review, published by the U.S. Naval War College Press, Newport, Rhode Island · Accepted pending revisions following peer review, August 2026 · Issue assignment confirmed September 2026 · Scheduled for the Winter 2027 issue

The Naval War College Review accepted the applicant’s manuscript on GOVERNANCE ENGINEERING for publication, and the Press has since confirmed that it remains in the editorial workflow and is scheduled for the Winter 2027 issue. The journal is the professional quarterly of the U.S. Naval War College, the senior professional military education institution of the United States Navy, and it is the venue in which American naval and joint strategic thought is customarily debated. The acceptance is the substantive point: the editor’s letter states that MOST OF THE PEER REVIEWERS’ COMMENTS CONCERN STRUCTURE AND THE NEED FOR BRIEF EXPLANATIONS OF CERTAIN TERMS RATHER THAN SUBSTANCE, which locates the required work in presentation rather than in the argument. Two substantive matters were raised. The first is terminological, a recommendation to replace one of the applicant’s terms with a different word for a class of systems, on the ground that his usage does not align precisely with current academic usage in cyberspace studies. The second is structural: reviewers questioned whether one engineering discipline should be identified as a primary lineage in its own right rather than treated as subsumed, with one reviewer recommending that his five-lineage construct become a six-lineage construct. The editor records that HIS ARGUMENT FOR SUBSUMING THE OTHER THREE FIELDS APPEARS TO HAVE PREVAILED, which is a notable detail, since it means the reviewers accepted the greater part of a taxonomy they were examining and dissented on one element of it. *** The subject is the discipline he has named. *** He defines governance engineering in his own published commentary as the design of mechanisms that specify, bound, monitor, degrade, transfer, recover and audit authority at runtime. A professional military journal accepting a manuscript under that heading, and debating the internal structure of its lineage argument, is evidence that the framing is being engaged as a field rather than as a proposal. *** This is recorded as an ACCEPTANCE and not as a publication. *** The manuscript is with the assigned associate editor for copyediting, citation validation and copyright review, revisions have not yet been returned, and the article is not counted toward the four peer-reviewed journal articles recorded elsewhere on this site. It will be moved when the issue appears. The editorial staff involved are not named in this record.

Peer-Reviewed Publication September 2026

Fourth peer-reviewed journal article published, in a fourth discipline

The ITEA Journal of Test and Evaluation · Volume 47, Issue 3 (September 2026) · ISSN 1054-0229 · Technical Articles section

The International Test and Evaluation Association published the applicant’s article “Bounded AI and a Layered Assured-PNT Architecture for Jamming-Resilient PNT” in volume 47, issue 3 of its journal. The article appears among the issue’s regular technical articles; the same issue also carries a separate section of papers from an annual defence and aerospace test and analysis workshop convened by the operational test and evaluation directorate, a national space agency, a federally funded research institute and the defence statistics section of a national statistical association. Contributors elsewhere in the issue are drawn from an Australian defence university, a national laboratory, a service academy, an applied physics laboratory and a university intelligence and security research laboratory. The journal operates a formal OPEN PEER-REVIEW model aligned to Committee on Publication Ethics principles, under which reviewers and authors are known to each other. The manuscript was assessed by two reviewers, the journal’s chief editor, who is a senior lecturer in test, evaluation and aircraft design at an Australian university, and an external research statistician, and it passed through TWO ROUNDS OF SUBSTANTIVE REVISION before acceptance. The first review required a new proof-of-concept test and evaluation section, engagement with three named recent publications, a feature-differential comparison table, and professional reformatting of thirteen tables and eight figures. A further editorial request added representative measures of effectiveness with illustrative threshold and objective values. The second review identified structural defects including unfilled subsections, an unfulfilled claim of an analytic error model, and inconsistent terminology. A final round of six editorial points required the applicant to soften several claims: a statement about the wider literature, a claim of an absence in the field, a comparative superlative, an assertion about failure containment, and a categorical conclusion about a technology’s readiness. Each was narrowed, and the novelty claim was reframed from a technology claim to an assurance-oriented synthesis. The article’s substance is a layered assured positioning, navigation and timing architecture combining bounded artificial intelligence, integrity monitoring and sensor fusion, in which authority boundaries are defined between machine recommendations, human decisions and deterministic decisions. Machine-learning authority is confined to advisory functions while consequential decisions remain under deterministic control and human oversight, which is the applicant’s capability-versus-authority separation applied to navigation resilience. The paper specifies a mixed software- and hardware-in-the-loop test campaign with the governor hosted on physically separate hardware so that the judged and judging mechanisms share no failure domain, traces each test back to a requirement or hazard, includes negative actuation tests against the trust boundary, incorporates human-autonomy teaming measures, and states plainly that the campaign is specified rather than performed. This is his fourth peer-reviewed journal article and his fourth discipline, after hardware-security engineering, security theory and digital war studies, and the first in test and evaluation, which is the field concerned with how a claim about a system is demonstrated rather than asserted.

View the issue ↗

Peer-Reviewed Publication August 2026

Third peer-reviewed journal article published, in a third discipline

Digital War (Springer Nature / Palgrave Macmillan) · Volume 7, article 6 (2026) · Published 31 August 2026 · DOI 10.1057/s42984-026-00113-1

Digital War, a Springer Nature journal published under the Palgrave Macmillan imprint in the field of digital conflict and war studies, published his article “The authority gap: why meaningful human control fails at the tier where it is signed” on 31 August 2026, ten days after acceptance, as volume 7, article 6 (2026), with a registered digital object identifier and his ORCID attached to the record. The article carries his Georgetown University affiliation, is designated an Original Paper, declares no conflict of interest, and cites twenty-two sources including work by the leading critical scholars of meaningful human control, United Nations expert-group papers on lethal autonomous weapons, and a General Assembly resolution. The significance for this record is threefold. First, it is his third article accepted by a peer-reviewed journal in three months. Second, it is his third distinct discipline, following a sensor-attestation protocol in a hardware-security engineering journal and a causal-mechanism specification in a security-theory journal, which establishes that the underlying research has been found publishable by reviewers in engineering, in social-scientific security studies, and now in war studies. Third, and most directly, the subject is his central claim rather than an application of it: where meaningful human control fails is a question about the tier at which authority is actually signed, which is the distinction between capability and authority that his architectures formalise and his deposited specifications mechanise. The applicant’s published commentary has argued that position across defence, engineering, scientific and policy venues throughout the year; its publication in a peer-reviewed war-studies journal places the same claim before referees in the discipline that owns the term. The article’s contribution is to explain a documented phenomenon rather than to restate it: a substantial critical literature has shown that human control over such systems is often closer to meaningless than meaningful, and he argues that this hollowness has an identifiable ARCHITECTURAL source. Authority is distributed across computational tiers, and the human signature attaches at a terminal tier downstream of the tiers where classification, proportionality and collateral-damage estimation are computed, so the endorser cannot review machine-paced, opaque reasoning within the decision window. He names the result the TERMINAL-SYSTEMIC AUTHORITY GAP, in which control is formally present yet substantively hollow. He distinguishes military decision systems from civilian algorithmic systems by development context, legal framework and deployment practice, reframes control as a property of the authority architecture governed across the full life cycle rather than as a moment of human endorsement, and draws implications for design, oversight and accountability while explicitly declining to assume that design alone can carry the burden. The keywords assigned to the published record include authority architecture and governance engineering, the terms he uses for his own research programme.

Read on SpringerLink ↗

Broadcast Expert Commentary August 2026

Live interview on France 24 English as an expert on autonomous weapons

France 24 English · Live from Washington · August 2026 · On-screen identification: “Burak Oktenli, Independent Researcher on AI Governance and Autonomous Systems”

France 24, the French state-funded international news broadcaster, invited the applicant onto its English-language channel for a live interview about a reported autonomous drone strike in Ukraine and what it indicates about the state of autonomous weapons. He appeared live from Washington opposite the studio anchor in Paris and was identified on screen by name and by the description independent researcher on AI governance and autonomous systems. This is a different class of recognition from the citations recorded elsewhere on this page: a broadcaster with a global audience selected him as the expert to explain a developing story in his field, in real time and without the opportunity to revise. His argument on air followed the position established across his published work. He held that the familiar demand for a human to remain in the loop is already obsolete for this class of weapon, since such systems navigate by onboard camera precisely because the radio link can be jammed, so the absence of a link is the design goal rather than a defect and there is no operator supervising the flight. From that he drew the consequence that the human decision cannot occur during the mission and must instead be made before launch, in the form of limits the machine carries with it. The limit he specified is an abort condition: if the integrity of the system’s sensors, or the evidence supporting its target, falls below the level that justified the mission, its permission to strike should lapse, the weapon should disarm, and the flight should end away from people. He was explicit that this cannot be a switch a person presses, because there may be nobody on the other end, and must instead be a boundary built into the machine beforehand, with the supporting evidence recorded so that anyone can afterwards reconstruct why the system was permitted to act. He added that the hardware is the more troubling element, since the module involved was a commercially available component costing a few hundred dollars rather than military equipment, and concluded that changing warfare does not require human-level artificial intelligence but only inexpensive machines performing narrow lethal tasks without supervision. The formulation he used to summarise the episode, that the machine did not become intelligent but that authority moved, is his research thesis compressed into a sentence, and the abort condition he described on air is the permission-lapse structure his architectures specify and his deposited formal models mechanise: authority that expires when the evidence supporting it fails, without requiring an operator to intervene, and leaves a record that can be reconstructed.

Independent Citation October 2026

Cited at the Modern War Institute at West Point by a retired Special Forces officer who advises the Army’s autonomy portfolio

Modern War Institute at West Point · Commentary, October 2026 · By a retired Army Special Forces lieutenant colonel, now an adviser and strategic integrator for the Army’s autonomy portfolio office · Names him and links his Modern War Institute commentary of June 2026

The Modern War Institute at West Point published a commentary, “We Shot the Wrong People: Autonomous Weapons and the Case for Human Judgment,” by a retired Army Special Forces lieutenant colonel, a West Point graduate with combat service in airborne, Ranger and Special Forces units who now advises the Army’s autonomy portfolio office. It argues that in close ground combat the decision to use lethal force must remain with a human who is meaningfully deciding. In the section setting out the way forward, the commentary names the applicant and links his own Modern War Institute commentary, “When the Machine Acts First: Closing the Authority Gap on the Autonomous Battlefield,” as a description of what governance looks like in practice: an authority enforcement layer that sits outside the artificial-intelligence system, pauses before irreversible action, and fails toward inaction when authorization is absent. *** It uses his design to show governance as engineering. *** The author argues that most debate about autonomous weapons concentrates on what a machine can perceive and do while treating oversight as an afterthought. Drawing on a four-layer view of intelligent systems in which governance is a layer of the system itself, and the layer most likely to be underbuilt, he defines governance as an engineering requirement: authority boundaries enforced at runtime, engagement envelopes the system cannot exceed, an abort capability it cannot override, and tamper-evident logging of every autonomous decision. The applicant’s published design is the example the commentary gives of that requirement in practice. *** Why it matters. *** The citation comes from a practitioner inside the field his research addresses: a former combat leader who now works on the Army’s acquisition of autonomous capabilities, writing in the West Point institute’s professional forum. It engages with the specific mechanism of the applicant’s work rather than mentioning it in passing. The author’s own arguments, including his five factors for deciding where human judgment must sit and his conditions for any offensive autonomous engagement, are his and are not attributed to the applicant.

Read the citing commentary ↗ His cited commentary ↗

Independent Citation August 2026

Cited and built upon in two federal rulemaking dockets by a runtime-governance software company

Runtime-governance software company · Reply comment NHTSA-2026-1552-0022, posted August 31, 2026 · Names his comment in the same docket (NHTSA-2026-1552-0010, posted August 25, 2026) and builds its central argument on his recommendation · Comment NHTSA-2026-1585-0009, posted the same day in a separate docket · Names his comment there (NHTSA-2026-1585-0005, posted August 25, 2026) and supports his recommendations

A software company that builds runtime governance for automated systems filed a reply comment in the federal vehicle-safety regulator’s docket on its automated-vehicle framework and interim guidance. The reply names the applicant’s comment in the same docket, filed six days earlier, among four it supports, and builds its central argument on his recommendation that applicants record data sufficient to reconstruct the authority state in force at any moment of operation. *** It sets his recommendation apart from the rest of the record. *** The company distinguishes claims that can be checked independently, such as a vehicle’s realized motion, from the authority under which an action was issued, which it describes as internal and leaving no physical trace. On that basis it concludes, in its own words, that his recommendation is not satisfied by a neighbouring commenter’s provenance proposal, because independent measurement corroborates what happened while a contemporaneous authority record is the only available evidence of what was allowed. *** It carries the recommendation into implementation. *** Noting that his results were demonstrated in simulation and formal verification rather than in deployed vehicles, the company reports having built a graduated-authority governor of its own, and a finding from it that bears on how his recommendation should be written: a per-decision record can be complete, signed, chained and self-verifying while still misstating the decision it documents, so an authority-state requirement should demand that sealed values be recomputable from the record and a published specification rather than merely retained. It offers draft guidance text to that end. *** It cites him again in a second rulemaking the same day. *** In a separate comment filed on August 31 in the regulator’s docket on an interim final rule amending its vehicle exemption procedures, the company names the applicant’s comment in that docket, also filed six days earlier, sets out his three recommendations for previously manufactured vehicles equipped with automated driving systems (that configuration conformance be the operative factor, that the operating record from before a grant travel with the vehicle, and that an entry record be a term of the grant), and states that it supports them, together with a state motor-vehicle administrators’ association’s recommendation on prior operating data. Reading the two together, it raises three questions it frames as its own: whether each submitted item was generated at the time of operation or reconstructed afterward, what retention period was actually in force and which vehicle-days left no surviving record, and what exposure basis supports any safety performance claim, since incident reports supply a numerator but not a denominator. It identifies his second recommendation as the one that would have disclosed that denominator, exposure by operating condition and supervision mode. *** Why it matters. *** Two filings on the public record of separate federal rulemakings, made the same day, that single out his recommendations, organise their arguments around them and, in the first, test them against a working implementation are direct technical engagement by a practitioner in the field his research addresses. The company states in both filings that it takes no position on any exemption application and makes no safety claim. Its positions, drafting language and implementation findings are its own and are not attributed to the applicant. Under the counting rule, the two filings are recorded as one citing source.

Read the reply comment ↗ His comment in the same docket ↗ Second comment, separate docket ↗ His comment in that docket ↗

Independent Citation September 2026

Translated in full into Russian across two instalments, with explanatory annotations added

Russian-language technology and broadcasting review · September 17, 2026 · Complete two-part translation of his Space Review analysis, with authorship, originating outlet and biography credited · FIFTEENTH COUNTRY in the citation record

A Russian-language technology and broadcasting review published a COMPLETE TRANSLATION of the applicant’s Space Review analysis on autonomous collision avoidance, issued in TWO INSTALMENTS, with his authorship and the originating outlet credited throughout and his biography carried at the end. *** A translation is a different order of engagement from a summary, and this one is unusually thorough. *** The publication did not extract a thesis or quote a passage; it rendered the argument in full, section by section, preserving his four rules, his three-layer separation, his position that a traffic-coordination service should remain an information service rather than a command authority, his statement that local optimisation is not automatically global safety, and his closing distinction that a collision estimate is data about a possible future while a manoeuvre is an intervention that alters the shared environment, between which an explicit grant of authority belongs. It also preserves his MANEUVER RECEIPT with its full field list. *** The editorial investment is the evidence. *** The translators added their own explanatory annotations for a readership outside the field, defining the probability-of-collision term, ephemerides, covariance, the traffic-coordination system and the policy directive behind it, and noting the membership of the space agency referenced. Adding an explanatory apparatus, across two instalments, for a technically specific argument by a foreign author is a level of effort no aggregator undertakes, and it indicates the editors judged the argument worth making legible to their audience rather than merely worth mentioning. Two further details are worth recording. His BIOGRAPHY IS CARRIED IN THE CORRECT FORM, stating that he holds an MBA and is pursuing a Master of Professional Studies, which is the accurate in-progress wording. And the reach is notable in itself: an argument written for an American space-policy readership about one operator’s regulatory filing was judged worth translating in full for a readership in a different country and language, on a publication whose ordinary subject matter is broadcasting and satellite equipment rather than autonomy governance. This entry is recorded from the second instalment, which the applicant supplied; the first is referenced by the publication as a companion. The site’s own editorial positions and annotations are its own and are not attributed to the applicant.

Read the translation, part two ↗

Independent Citation September 2026

Reconstructed at length in Chinese by a service that explains foreign analysis to a Chinese-reading audience

Chinese-language analysis service tracking overseas commentary on China · September 14, 2026 · A dedicated article reconstructing his argument, with attribution throughout and a link to the source

A Chinese-language analysis service that tracks how foreign commentators write about China devoted a full article to the applicant’s Washington Examiner op-ed on artificial-intelligence diplomacy toward developing countries, naming him throughout, attributing positions to him directly, and linking to the original. *** This is his first treatment in Chinese, and the first instance of his argument being carried back into the language of the country it concerns. *** The service is not a wire aggregator. It maintains keyword taxonomies, a source index and a published editorial method, and its business is explaining foreign analysis to a Chinese-reading audience. Its article runs to several sections, reconstructs his reasoning in sequence rather than excerpting it, and marks his conclusions explicitly as his, using formulations equivalent to his core judgment is and according to the author. *** What made the piece citable for that audience is the even-handedness of the original, and the summary preserves it. *** It records him crediting the new organisation’s offer as substantive rather than symbolic, responding to real capability gaps in training, cooperation centres, warning systems and downloadable models, and it carries his reciprocal principle in full: that a safety standard observed only inside one technical bloc remains a bloc standard, and that AN INSTITUTION LED BY THE OTHER SIDE WHICH EXCLUDED THIS ONE COULD EQUALLY NOT CLAIM TO GOVERN GLOBAL ARTIFICIAL INTELLIGENCE. It also preserves his refusal of the convenient framing, recording his position that the contrast between closed and open model provision is A SNAPSHOT OF COMMERCIAL ARRANGEMENTS RATHER THAN A CIVILISATIONAL ATTRIBUTE, that open weights are not the same as open source, and that demanding developing countries refuse useful technology is not a strategy. That a service writing for a Chinese-reading readership chose to reconstruct an American commentator’s argument at length, including the passages least flattering to its own side and the passages least flattering to his, is a stronger indicator of the argument’s analytical standing than agreement from a sympathetic outlet would be. The applicant’s op-ed itself sits behind a subscription wall, so this summary is an independent party’s characterisation and is recorded as such; the record of his own piece elsewhere on this site remains held to what the outlet makes publicly accessible.

Read the Chinese-language treatment ↗

Independent Citation June 2026

Curated by an Indo-Pacific strategy analysis service, which adopted his own term in its headline

Independent geopolitics, cyber and strategy analysis service with an Indo-Pacific focus, operating a network of seven regional and sectoral titles · June 22, 2026 · Editorial summary with attribution and source link

An independent analysis service covering geopolitics, cyber and strategy for an Indo-Pacific readership selected the applicant’s Space Review article on autonomy in space for its news record, summarising the argument in its own editorial voice, naming him as its author, and linking to the source. *** The detail worth recording is the headline. *** The service did not reuse the applicant’s title. It wrote its own, and that headline states that space autonomy NEEDS AN AUTHORITY ARCHITECTURE, which is his own term adopted by a third party writing independently for its own audience. Its summary characterises the argument accurately, that growing threats to space assets mean greater reliance on autonomy, and that this requires deciding now WHAT MEASURES SPACECRAFT MAY TAKE AUTONOMOUSLY AND WHO WILL BE RESPONSIBLE FOR THEIR EFFECTS, which is the applicant’s distinction between capability and answerability stated by someone else. The weight of this entry is stated plainly rather than inflated. It is a CURATED EDITORIAL SUMMARY AND NOT AN ANALYTICAL ENGAGEMENT, so it carries less evidentiary weight than the citations on this page in which an author examines, tests or disputes his reasoning. What it does establish is selection and characterisation by an independent specialist service that maintains published methodology, corrections and authorship pages, and that its editors found the argument worth carrying to a regional strategy audience under a heading drawn from his own vocabulary.

View the record ↗

Independent Citation September 2026

Engaged at length by an independent defence commentator, who credits the analysis and challenges its scope

Long-form critique on an independent defence and national-security newsletter · September 14, 2026 · A substantive and MIXED assessment, recorded here with its criticisms intact

An independent defence commentator published a detailed critique of the same national science and technology strategy the applicant had analysed for RealClearDefense, opening with the applicant’s piece and treating it as the reading that introduced the strategy to that readership. *** The credit is specific. *** The author writes that the applicant’s central move, arguing that an autonomous system’s permission to act should CONTRACT AS THE EVIDENCE UNDERNEATH IT DEGRADES rather than treating a stable confidence score as licence to keep operating, is genuinely sharp analysis and sharper than most of what he receives on artificial-intelligence governance. He separately records that the autonomy-governance angle is real, worth writing about, and worth crediting the applicant for SURFACING BEFORE MOST OF THE TRADE PRESS caught up to it. *** The criticisms are recorded here in full, because a citation reported selectively is worth less than one reported accurately. *** The author makes two. First, that the strategy’s graceful-degradation language appears in a pillar describing physical infrastructure resilience, and that the applicant took a hardware-redundancy concept and extended it into a permissions-and-authority concept for autonomous systems; he calls that a LEGITIMATE ANALYTICAL MOVE while holding that it is not what the document proposes on its own terms, and that presenting an extension as the strategy’s explicit argument is a seam a careful reader should keep visible. Second, that the piece treats one of twelve designated technology areas as the strategy’s centre of gravity, which he describes as a keyhole view standing in for the whole room, and argues the document’s more revealing feature is that it is precise where precision serves enforcement and vague where vagueness protects it from being checked. On the first criticism, the record permits a reader to judge for themselves. The applicant’s published text states that the strategy treats resilience as a property of hardware, networks and supply chains while STOPPING SHORT OF AUTONOMY ITSELF, and presents his authority-degradation test article as what should follow, which is the structure of an extension rather than a claim about the document’s own argument. Whether that is stated plainly enough is the reviewer’s point and is left standing here. The entry is recorded as an independent citation because it is exactly that: an unsolicited, dated, substantive engagement by a third party who read the work closely enough to distinguish the applicant’s contribution from his source, and who credits the contribution while disputing its framing. Sustained critical engagement of this kind is a stronger indicator of a work being taken seriously than agreement is.

Read the critique ↗

Independent Citation September 2026

Cited by name in a competing essay, which concedes that his argument arrived first

Long-form essay on agentic artificial intelligence and financial regulation, by a named independent author who also builds and sells systems of the kind the essay advocates · September 8, 2026 · Cited in the essay’s section of objections to its own thesis

An independent author published a long-form essay arguing that a deterministic, pre-execution refusal mechanism for agentic systems is buildable, taking as its evidence a United States market-access rule that has required automated pre-trade controls on every order since 2011. In the essay’s sixth section, titled as the case against its own argument, he addresses the applicant directly. *** He concedes priority. *** He records that the applicant published a piece four days before his own essay was drafted, making a closely related argument from the same market-access rule together with adjacent European and United Kingdom algorithmic-trading controls, and summarises the applicant’s position accurately: that an artificial-intelligence agent should never hold one broad permission to trade or act, and that those rules model the narrow, revocable authority an agent should instead be scoped to. His assessment of it is unambiguous, that this is a real argument and that IT ARRIVED FIRST. He then distinguishes his own claim as differing in kind rather than emphasis, being about whether a deterministic mechanism can refuse a specific action before it executes rather than about how much authority an agent should hold, states that THE TWO ARGUMENTS ARE COMPATIBLE AND SHOULD PROBABLY BE READ TOGETHER, and adds that neither author may claim the underlying analogy as original since that credit belongs to the rulebook. Three features make this stronger than an ordinary citation. The acknowledgement appears in a section the author built to state the strongest objections to his own essay, so the applicant’s work is treated as a serious competing position rather than as supporting material. The author declares in the same essay that he sells a product of the kind it advocates and asks readers to weigh his argument accordingly, which makes his concession of priority to an unaffiliated researcher a statement against his own commercial interest. And the citation is specific and checkable, naming the applicant, the outlet, the publication date and the substance of the argument, so a reader can verify the claim of precedence independently. The piece cited is the applicant’s analysis of per-action-class authority in financial systems, published on 4 September 2026 and listed separately in the publications record.

Read the essay ↗

Independent Citation September 2026

Cited and corroborated by an experimental cosmologist against his own field’s canonical case

Newsletter of a university physics professor and principal investigator, distributed to roughly 3,500 subscribers · September 2, 2026 · Featured item · Also republished by the author to a general-readership platform

An experimental cosmologist who directs a research laboratory, hosts a long-running science interview programme, and has written a book on premature discovery claims devoted a featured section of his newsletter to the applicant’s argument. He summarised it as the position that scientific discovery should be treated as a rank earned through evidence rather than as excitement or statistical significance alone, which is the thesis of the applicant’s Eurasia Review analysis on that subject. What makes this more than a citation is what he did next. Rather than simply endorsing the argument, he tested it against the canonical case in his own discipline: a 2014 result in which an excess polarisation signal in the cosmic microwave background was initially interpreted as evidence of primordial gravitational waves from cosmic inflation. He notes that the signal itself was real, that galactic dust could produce a similar pattern, and that later joint analysis with data from another instrument showed dust explained much of it, leaving no statistically significant detection. From that he draws the lesson in the applicant’s own terms, that a real anomaly does not automatically justify a discovery claim, and that instrument validity, confounders, statistical treatment, alternative models and replication must all be checked. He then records that the applicant extends the warning to artificial intelligence, which can rapidly generate confident interpretations before the evidence chain is complete. This is corroboration by a practitioner with direct experience of the failure mode being described, in a field where that episode is the standard teaching example, and it is the first recorded instance of a working experimental scientist validating the applicant’s inference argument against a case from his own research programme. The characterisation of the argument recorded here is the cosmologist’s own summary; the applicant’s underlying analysis is listed separately in the publications record.

Read the newsletter ↗ Republished version ↗

Academic Citation August 2026

Cited in a peer-reviewed journal article published by Wiley

Middle East Policy · Original Article · 2026 · DOI 10.1111/mepo.70106 · Endnote 43

A peer-reviewed article in Middle East Policy, published by Wiley on behalf of the Middle East Policy Council, cites the applicant’s Geopolitical Monitor analysis “The Reciprocity Inversion: How Ukraine Became the Gulf’s Counter-Drone Lifeline” at endnote 43. The citation supports a substantive factual claim in the body of the article rather than a passing reference: the authors state that emerging partnerships have acquired a longer-term and reciprocal character, with ten-year security agreements concluded with two Gulf states in which advanced air-defence missiles are sought in exchange for counter-drone expertise, and attribute that reading to the applicant’s piece. It sits in a sequence of endnotes drawn otherwise from wire services, think-tank research and defence trade press, and it is the source carrying the reciprocity argument that the surrounding sentence advances. The authors are an artificial-intelligence and technology initiatives librarian at a United States research university and a professor of security studies at a European university, and the work is supported by a European research fund. This is the applicant’s first recorded citation in a peer-reviewed journal article issued by a major academic publisher, as distinct from the preprints, student-edited journal and government record documented elsewhere on this page.

Citing article (DOI) ↗ The cited analysis ↗

Independent Citation August 2026

Quoted on where responsibility for autonomous systems properly sits, in a syndicated feature

TomDispatch feature by a bylined contributing writer · Carried by six publications between August 26 and September 2026: The Intercept, CounterPunch, Popular Resistance, ZNetwork, Foreign Policy In Focus, and The Smirking Chimp

A bylined feature on the governance of autonomous weapons, distributed by the syndication service TomDispatch and carried by The Intercept and subsequently by CounterPunch, Popular Resistance, ZNetwork, the Institute for Policy Studies project Foreign Policy In Focus, and The Smirking Chimp, cited and directly quoted the applicant’s Articles of War analysis. The article identifies him by name as an independent researcher on artificial-intelligence systems and their governance, names the publication in which the analysis appeared, and uses his argument as the authority for the central legal question the feature raises: where responsibility lies for the actions of an autonomous system when no individual operator directed a particular engagement. It reports his position that existing law continues to apply when machines make battlefield decisions, and quotes him directly to the effect that such law must “be applied earlier, at the point where the architecture of a swarm is chosen, rather than after a strike when the question of responsibility has already become difficult to answer.” The feature also sets the underlying proposition in display type, stating that true responsibility for the actions of an autonomous swarm is properly located much earlier in the sequence, at the point where such a system is first conceived and designed. This is the applicant’s architectural-accountability argument used as load-bearing authority rather than as illustration, and its syndication across more than one general-readership publication is the widest circulation his legal analysis has received. It is recorded as a single citation carried by multiple outlets rather than as multiple citations, since one writer quoted him once and the syndication service distributed that article onward. This record notes the citation and the argument quoted; it takes no position on the wider contents, characterisations, or editorial judgements of the article in which the citation appears.

The Intercept ↗ CounterPunch ↗ Popular Resistance ↗ ZNetwork ↗ Foreign Policy In Focus ↗ The Smirking Chimp ↗

Independent Recognition July 2026

Recommended to a professional readership by a law-of-armed-conflict scholar

Public note by a United States law professor who directs an international and comparative law centre and publishes on the law governing armed conflict · July 26, 2026

A United States law professor specialising in the international law of armed conflict, who directs an international and comparative law centre and writes a public commentary series on legal responses to crisis, posted a note recommending the applicant’s Articles of War analysis to his readership. He described it as an excellent piece on the accountability problems that arise under international humanitarian law from the design choices and deployment methods of drone swarms in armed conflict, and summarised its structure accurately, noting that the applicant examines three distinct ways swarms have been deployed and how each affects the difficulty of making after-the-fact determinations about human judgment and accountability. The recommendation is unsolicited, made under the scholar’s own name on a public and dated platform, and demonstrates engagement with the argument rather than mere circulation of a link, since it restates the analytical structure of the piece correctly. It is recorded here as a professional recommendation rather than a formal scholarly citation, since it is not a reference in a peer-reviewed work and the site does not present it as one. Its weight lies in the reviewer’s standing: the applicant’s piece appeared in the legal-analysis journal of a military academy institute and turns on questions of international humanitarian law, and the person recommending it teaches and publishes in exactly that field.

Read the note ↗

Independent Citation June to September 2026

Selected eight times, five with original analytical commentary, by an Indian strategic-studies repository

Indian Strategic Studies · Curated research portal on military strategy, multi-domain operations, cyber warfare and open-source intelligence · Eight selections between June and September 2026

Indian Strategic Studies, a curated digital repository and research portal maintained by a retired Indian Army major general who formerly served as senior directing staff at his country’s National Defence College and specialises in electronic warfare, signals intelligence, cyber and communications, has selected eight of the applicant’s pieces between June and September 2026, each carried under a credit line naming the original outlet and the author. The first three were accompanied by editorial summaries written by the curator. The five most recent, two of them carried within five minutes of each other on the same morning, go further and add original analytical commentary extending the applicant’s arguments, followed by a strategic question posed to the portal’s readership and the curator’s own assessment in reply. On the navigation-integrity analysis he reasoned that separating verification of signal origin from validity of the resulting position shifts the bottleneck in automated decision loops from signal processing to the delegation of command authority, that automated systems lacking defined confidence-degradation parameters will either fall back on last-known values or suspend sensor fusion entirely, and that command architectures receiving low-latency updates without explicit integrity flags are exposed to automation bias, concluding that without system-enforced degraded modes commanders will absorb navigation uncertainty rather than interrupt tempo. On the testing analysis he reasoned that delegating terminal guidance to onboard processing upon link degradation changes the engagement timing and spatial boundaries defenders must plan against, and that evaluating such hybrid control loops requires multi-spectral stress testing across degraded navigation and visual deception rather than the static electromagnetic thresholds by which proving grounds have traditionally measured success, concluding that raw interference power will yield to rapid sensor fusion at the tactical edge. On the machine-speed escalation analysis he reasoned that embedding automated decision engines into existing autonomy policy shifts the operational locus from deliberate human planning toward parameters fixed in software beforehand, that sub-second sensor processing leaves no practical window for a commander to review a target validation, and that authority therefore migrates from the officers presiding over operations to those who configured the engagement logic, concluding that without mandatory latency buffers the ability of senior commanders to halt an automated sequence erodes structurally. On the cyber-resilience analysis he reasoned that where an intrusion reaches low-level firmware or cryptographic roots the standard incident-response protocol becomes self-defeating, since the attacker has compromised the attestation mechanism itself rather than merely extracted data, and that update channels can become distribution vectors for the very thing they are meant to remedy; he then put the trade-off to readers as a choice between delayed containment under human authorisation loops and automated over-isolation of critical command nodes, assessed that uncontrolled autonomous remediation is the more severe systemic risk because it can sever command links instantly, and concluded that strictly bounded authority envelopes will be necessary to prevent automated defence software from causing unintended disruption. That conclusion adopts the applicant’s own construct and requirement as the curator’s analytical answer. On the target-data analysis he reasoned that automated processing compresses the sensor-to-shooter cycle while stripping away the time buffers that once allowed analysts to verify that a target was still valid, and that the remedy is to build data-decay measures into the targeting architecture itself, treating target records as perishable assets with distinct half-lives so that automated engagement recommendations are restricted when supporting sensor feeds lapse; he then asked readers how commanders can set verifiable thresholds for data decay without sacrificing tempo, and assessed that automated pauses on conflicting sensor data will slow operations only when targets are volatile, preserving speed against fixed infrastructure while enforcing caution against fluid target sets. The half-life framing is the applicant’s own, taken from his argument that the elements of a target record expire at different rates. A curator returning to an author’s work eight times, writing original analysis on top of it, and inviting a professional readership to debate the result, is a substantially stronger form of engagement than selection alone.

Most recent feature ↗

Independent Citation August 2026

Featured by the Resilient Navigation and Timing Foundation

Resilient Navigation and Timing Foundation · Blog · August 21, 2026

The Resilient Navigation and Timing Foundation, a United States scientific and educational charity that advocates on positioning, navigation and timing resilience and convenes an international advisory council, corporate members and an annual leadership summit, featured the applicant’s RealClearDefense analysis on its blog the day after publication. The foundation’s editor introduced it under the publication’s standard framing, describing it as an article about authentication and confidence in navigation and timing solutions and adding that many factors go into the safe and resilient use of such signals, so that ignoring or misjudging one could produce disaster. The item is noted as having been brought to the foundation’s attention by its own readers, which indicates the analysis circulated within the professional community before the foundation carried it. The feature places the applicant’s argument before the specialist constituency that sets and consumes navigation-resilience requirements, alongside the foundation’s coverage of aviation-safety interference bulletins, national receiver programmes and satellite-navigation policy. The foundation’s page headline attributes the article to a different defence publication; the piece was in fact published by RealClearDefense, whose author page the foundation’s own byline link correctly points to, and this record follows the canonical source.

Read the feature ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 10, 2026 · Authored by Burak Oktenli · Eighteenth placement with the outlet · Second piece published that day

The Washington Examiner published the applicant’s op-ed “SpaceX satellites dodge collisions 1,100 times a day. Who’s actually driving?,” his eighteenth placement with the outlet and the second he published that day. It moves his authority question into civil space operations, and the reasoning in its opening is the part worth recording. *** He declines the available alarm. *** The piece opens on a published figure for propulsive collision-avoidance manoeuvres performed by one operator’s fleet over six months, amounting to more than a thousand a day, and his first analytical move is to say what the number is not: it is not evidence of recklessness. He attributes it instead to a conservative safety threshold, the sheer size of the fleet, and an unusually transparent public accounting, and the last of those is a point in the operator’s favour, since the figure is known only because it was disclosed. Having removed the sensational reading, he states the reading that survives: the volume is a policy alarm bell rather than a safety indictment, because at that scale A RULE INSIDE SOFTWARE IS NO LONGER MERELY ADVICE TO AN OPERATOR, IT BECOMES PHYSICAL BEHAVIOUR IN ORBIT. The question he then puts is the one his research programme asks of every autonomous system, transposed to a civil domain: what turns a collision warning into PERMISSION TO MOVE A SPACECRAFT. That is the separation between an estimate and an authority to act on it, which his formal work treats as the distinction between what a system can compute and what it is permitted to do, and his argument is that the decision should be made before American satellite fleets grow more autonomous rather than after an event forces it. The framing carries his standing position that a risk score is an input to an authority decision and not the decision itself, and that the threshold converting one into the other is a governance choice which should be explicit and owned rather than embedded silently in software. The body of the piece sits behind the outlet’s subscription wall and is not recorded here; this entry is held to the argument in the accessible portion, which comprises the framing, the declined inference and the question posed. No operator, agency, regulator or programme is named in this record, and no figure beyond the published maneuver count is attributed.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 10, 2026 · Authored by Burak Oktenli · Seventeenth placement with the outlet

The Washington Examiner published the applicant’s op-ed “Trump is right about quantum. Now the Pentagon has to prove it works,” his seventeenth placement with the outlet. It responds to a newly finalised federal programme taking minority equity stakes in three quantum computing firms alongside research and development support, and it follows the pattern that characterises his policy writing: it grants the premise, then identifies the specific risk the premise creates. *** He supports the policy before qualifying it. *** He calls it a serious use of industrial policy, accepts that the technology could affect secure communications, navigation without satellite positioning, sensing, materials and computing, agrees the capacity should be built before it is urgently needed, and states that taking equity is more disciplined than writing a grant with no claim on future value. The risk he then names is precise and follows directly from the policy rather than opposing it: GOVERNMENT OWNERSHIP CAN BE MISTAKEN FOR GOVERNMENT VALIDATION. An award may help a company solve a manufacturing bottleneck without establishing that its technology delivers a military advantage, and the proposed remedy is that the defence department protect the investment by acting as a demanding customer, under which every quantum claim competes against the best available alternative under realistic conditions regardless of which supplier holds federal backing. His central analytical move is to reject the single-race framing. For acquisition purposes the field runs on AT LEAST THREE CLOCKS: post-quantum cryptography, which is ready for implementation and runs on classical computers, so procurement need not wait for a quantum machine but still requires evidence that products implement the standards correctly, manage keys securely, interoperate with legacy systems and can be upgraded, since standardised mathematics does not make every implementation trustworthy; quantum sensing and photonics, moving through prototypes and platform integration, where he praises the modest scale of a service feasibility competition as a virtue because it exposes technical weaknesses before a laboratory result becomes an expensive programme, and notes a research agency programme requiring compact sensors to be tested on a government-provided airborne platform where vibration, interference and field gradients punish fragile equipment; and general-purpose quantum computing, on the longest path, where a benchmarking initiative defines utility in economic terms and has staged companies through risk-reduction planning into government verification and validation. A single standard of quantum readiness, he argues, would confuse urgency with maturity. The whole-system rule is his standing requirement restated for this domain: the stopwatch must include data loading, state preparation, error correction or mitigation, measurement, classical processing, power, cooling, specialised infrastructure and operating cost, and the comparison must use a TUNED CLASSICAL SYSTEM on the same task, or the department could buy an impressive calculation whose advantage disappears when the rest of the mission returns. His proposed instrument is a ONE-PAGE QUANTUM EVIDENCE RECORD stating the mission task, the strongest non-quantum comparator, the whole-system measure of success, the conditions actually tested, the principal failure modes, and the next experiment capable of disproving the claim, attached to existing technical reviews, and he is explicit that this is an extension of an established research agency’s questions rather than a new approval board. He then adds the requirement that connects this to his research programme: where a quantum result will influence an operational decision, the record should define WHAT WEIGHT THE RESULT RECEIVES OUTSIDE ITS TESTED CONDITIONS, so that a navigation sensor may deserve greater trust after surviving a denied-positioning flight test but not beyond its demonstrated drift, vibration or calibration limits, and a computing result may inform planning without becoming the sole basis for a consequential decision. His formulation, that AUTHORITY SHOULD FOLLOW EVIDENCE, NOT BRANDING, is the capability-versus-authority separation applied to procurement. He supports it by noting that a national security agency already applies the same logic to quantum communications, favouring post-quantum algorithms for national security systems while treating distribution as highly dependent on implementation, and evaluating the complete security system rather than awarding credit for the word quantum. His conclusion addresses the conflict the equity stakes create. Companies receiving federal capital should remain eligible to win defence work, but the government’s financial interest cannot become a shortcut through acquisition gates, and vendors without federal equity should be allowed to beat portfolio companies on performance. He frames common tests as protecting taxpayers, warfighters and credible quantum firms alike from the reputational damage caused by exaggerated claims, and closes on the line the piece is built toward: taxpayers should share the upside, and warfighters should receive only the advantage that survives the test.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 10, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “The Autonomous Weapons Agreement Will Be Decided in Export Markets,” his sixty-first contribution to the outlet. It follows his earlier piece on treaty verification and answers the question that one leaves open: where a non-binding consensus acquires practical force. His argument is that it acquires it in procurement and export licensing rather than in treaty text, and the reason he gives is structural. A PLATFORM CROSSES A BORDER ONCE; ITS AUTONOMY CAN CROSS REPEATEDLY. A new model, retrained data set, targeting application, cloud service or software patch can alter the behaviour of a fielded system without a new airframe leaving a factory, so a licence can remain attached to hardware while the capability that justified it changes around it. He notes that export-control systems were not designed for that fluidity, that intangible transfers are less visible to licensing authorities, and that a stand-alone transfer provision present in an earlier negotiating text did not survive into the draft circulated before consensus, leaving an implementation gap rather than a legal permission. He is careful about what a national legal review can settle. An exporting state’s review cannot answer the importing state’s question, because the buyer may use different mission data, doctrine, environments, target classes and supporting systems, so a configuration judged lawful in one setting does not settle the review in another. His four proposed provisions follow from that: a PRECISE DELIVERED BASELINE identifying the software, models and data dependencies forming the accepted system and which components remain under the supplier’s remote control, which he notes is ordinary configuration management applied to a legal problem and requires no disclosure of source code; a definition of MATERIAL CHANGE triggering notice and renewed review before operational acceptance, covering changes delivered as a subscription or cloud service; BUYER ACCESS sufficient for an independent legal and operational assessment, with sensitive evidence held in protected annexes, the objective being comparable review rather than public disclosure; and POST-SALE SAFETY REPORTING in both directions, preserved through upgrades, resale and integration into another platform. He states the limits of his own proposal, that contractual implementation cannot bind non-participants, prohibit every unacceptable system or resolve disputes over the proper scope of autonomy, and its value, that it can shape the installed base while formal negotiation continues, since weapons procured now may serve for decades while their software changes many times. His argument to suppliers is commercial rather than moral: a producer able to deliver stable baselines, clear change notices and review-ready evidence reduces the customer’s legal and integration risk, so responsible autonomy becomes a feature that helps a system enter trusted markets rather than a restriction added afterwards.

Read on Eurasia Review ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 9, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “A Four-Hour Fault, a Two-Day Crisis: Britain Needs a New Measure of Air Traffic Resilience” under its Analysis designation, his sixtieth contribution to the outlet and the twenty-fourth to carry that label. *** He declines to diagnose before the evidence arrives. *** He writes that it is too early to say why the system failed or whether the fault was preventable, records the operator’s statements that safety was not compromised and that a cyberattack was not the cause, and states that a breaking-news failure is an invitation to investigate rather than a licence to assign negligence. The judgement he does make is definitional: the initiating event and the harm experienced by the network are DIFFERENT OBJECTS, and SAFE FAILURE IS THE FLOOR OF RESILIENCE, NOT ITS FULL DEFINITION. The useful question is therefore not why the computer failed but how well the whole system absorbed the failure. His measurement proposal follows from that separation. A published RECOVERY CURVE plotting the share of scheduled capacity actually delivered against time since the first restriction, where the area between planned and delivered service represents the disruption users experienced and a long shallow tail reveals fragility that a system-restored timestamp conceals. TWO CLOCKS in every major incident report, one recording technical restoration and the other network recovery, with neither permitted to stand in for the other. An ATTRIBUTABLE-CASCADE MODEL separating direct effects, materially amplified effects and unrelated disruption, so that the public cost does not disappear between institutional boundaries and the operator is not charged for unrelated weather or airline constraints. And a MINIMUM SERVICE ENVELOPE defined in advance for major failure classes, making degraded performance a designed operating state rather than an improvised interval between normal service and shutdown. He is careful about a prior remediation programme, noting that a completed set of recommendations from an earlier incident does not establish resilience across unfamiliar failure classes, and that the test of learning is not whether the last fault can recur in the same form but whether the next fault has a smaller blast radius. His strategic observation is that the ruling out of hostile action sharpens rather than ends the lesson: if a benign failure can create a national event, an adversary need not invent a more spectacular outcome but only find another route to the same dependency, which places operational resilience beside cybersecurity in the protection of critical infrastructure.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 8, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Faster AI Is Not Always Better Astronomy,” his fifty-sixth contribution to the outlet and the fourteenth in a sequence on machine-assisted scientific inference. Its thesis is that detection software should be judged by THE EVIDENCE IT PRESERVES as well as the events it finds, and the distinction it rests on is precise: a missed preservation trigger differs in kind from a classification error, because a label can be corrected while the underlying record survives, whereas information the instrument never retained cannot be recovered from a summary. An algorithm can therefore improve a speed benchmark while narrowing the universe future researchers are able to study. *** The piece reports his own negative result and declines to explain it away. *** He describes a deliberately limited pilot comparison in which an AI-guided route evaluated about a quarter of the search grid used by a classical baseline, which appeared promising as a reduction in arithmetic, and then reports that the baseline recovered more of the blinded injections than the AI route, that the AI route missed cases the baseline caught while finding none the baseline missed, and that it took longer overall despite performing fewer calculations. His conclusion on those numbers is stated plainly: the proposed replacement had not earned deployment. He then states the limits of his own test as carefully as its findings, recording that it involved one instrument, one source, short processed records and a simple baseline, that the difference does not establish a general disadvantage, and that the timing result is specific to that implementation, adding that a negative result is useful when the test is allowed to change the decision. He also requires that any proposed acceleration be compared against the best relevant conventional method with data movement and verification included, since beating a weak implementation is not evidence of progress, and extends the same accounting to future quantum proposals. The governance requirements he proposes are his standing ones in a scientific register: treat detection software as part of the instrument, since a model update that changes which events are retained changes the survey’s view of the sky; keep a conventional path that may fail differently; run continuing injection tests and record every model and threshold change; provide a fallback when inputs move beyond validated conditions; keep model-generated reconstructions distinguishable from recorded data, since a plausible-looking feature is not an independent measurement; and reserve capacity for an explicit UNKNOWN, PRESERVE state together with a randomised sample of windows that did not trigger retention, on the ground that a system auditing only its accepted candidates cannot assess its blind spots. His closing formulation is that the most consequential error may be a strange, low-confidence event quietly denied a place in the permanent record, since no future algorithm can reanalyse a measurement that was overwritten.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 19, 2026 · Authored by Burak Oktenli · First piece published that day · The general-domain statement of his orbital collision-avoidance argument

Eurasia Review published the applicant’s opinion piece “An AI Warning Is Not Permission to Act,” his eighty-first contribution to the outlet and the first he published that date. *** The title is constructed identically to the one he published five days earlier in a space-policy journal about orbital collision avoidance, and this is the general-domain statement of that argument. *** Where the earlier piece held that a modelled collision probability is evidence about a possible future rather than permission to change a shared environment, this holds the same of any machine-generated warning in any institution, and the parallel establishes the claim as structural rather than domain-specific. He separates THREE QUESTIONS that institutions collapse into one score. ATTENTION asks why somebody should look at this now. EVIDENCE asks what is actually known, what is missing, and which alternative explanations remain plausible. AUTHORITY asks who may act, under which rules, with what reversibility and accountability. His account of how they merge is precise and is the observation the piece turns on: a score turns red, a label appears, a response button sits beside it, and the operator experiences a progression that feels natural, URGENT THEREFORE LIKELY, LIKELY THEREFORE ACTIONABLE. His conclusion is that MEANINGFUL HUMAN CONTROL CAN DISAPPEAR WITHOUT ANYONE FORMALLY REMOVING THE HUMAN, which is his standing argument that a person who cannot interrogate what they are approving is not exercising control. *** He cites evidence against the intuition that pairing a person with a model improves the decision ***, recording that research on automation bias shows people can over-rely on incorrect machine advice and that explanations do not necessarily remove the tendency, and that a recent meta-analysis found human and machine combinations performed worse on average than the better of the human or the model alone in decision tasks. That is a finding unhelpful to the reassurance most institutions offer, and he uses it rather than avoiding it. His governing rule is that URGENCY SHOULD RAISE ATTENTION WITHOUT RAISING AUTHORITY, illustrated with a case in which both statements are simultaneously true: a low-confidence report of a potentially catastrophic compromise demands immediate investigation, while the same uncertainty makes automatic isolation of a critical network inappropriate because the intervention itself could interrupt essential operations. He then extends the point to the case where THE MODEL IS CORRECT, since a system may accurately identify unusual financial behaviour without evidence of fraud, correctly estimate elevated clinical risk without knowing which intervention suits a particular patient, or correctly identify anomalous traffic while lacking the context to distinguish attack from configuration error. The two sentences he draws from it are the compact form of his programme: ACCURACY DOES NOT CREATE JURISDICTION, CONFIDENCE DOES NOT CREATE PERMISSION. He is careful to distinguish this from adjacent debates he has also written about, noting that stop rules govern whether a capability advances, kill switches govern whether a system continues operating, and RESPONSE ARCHITECTURE governs what an institution may do with a machine-generated warning. His three gates follow: an ATTENTION gate that preserves urgent warnings even when evidence is weak, whose job is triage; an EVIDENCE gate exposing source quality, missing information, alternative explanations, model limits and relevant uncertainty, so that the reviewer INHERITS AN EVIDENTIARY DOSSIER RATHER THAN A CONFIDENCE BAR; and an AUTHORITY gate specifying which actions are permitted at that evidence level, who owns the decision, whether the action is reversible and when a second approval is required. He states the proportionality limit himself, that most low-consequence reversible automation should remain fast and that FRICTION BELONGS WHERE THE COST OF UNJUSTIFIED ACTION BECOMES SERIOUS. The design consequence he draws is one an interface designer can act on: a system needs to make FOUR responses possible, to intervene, to escalate to a qualified human, to gather more evidence, or to DELIBERATELY DO NOTHING YET, and if the interface makes only approval feel like action then the response architecture is already biased. He carries this into procurement, arguing that buyers of consequential decision-support should require vendors to demonstrate that an ALERT CAN REMAIN VISIBLE EVEN WHEN AUTOMATED ACTION IS BLOCKED, with audit records able to reconstruct what the model observed, what it inferred, what evidence was missing, what it recommended, who had authority to approve and what was actually done. *** He then states the limit of his own discipline, which is the most self-restrained passage in his published record. *** None of this, he writes, makes an institution’s rules legitimate: A PERFECT AUDIT TRAIL CAN DOCUMENT A BAD POLICY, an authenticated official can exercise authority badly, and ENGINEERING CAN PRESERVE THE CONDITIONS FOR ACCOUNTABILITY BUT CANNOT MANUFACTURE LEGITIMACY. A researcher whose programme is the engineering of accountability naming the boundary beyond which that engineering delivers nothing is a form of restraint that strengthens every claim he makes inside it. His closing three-line statement is the clearest summary of the distinction his work rests on: a warning is a claim on attention, evidence is a claim on belief, authority is a claim on action, and a system SHOULD NEVER RECEIVE THE THIRD SIMPLY BECAUSE IT WAS GOOD AT EARNING THE FIRST. No companies, laboratories, governments, standards bodies or figures are attributed in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 18, 2026 · Authored by Burak Oktenli · Second of his pieces on timekeeping as evidence

Eurasia Review published the applicant’s opinion piece “The Race Beyond GPS Needs a Rule for When Machines Stop Trusting Time,” his seventy-eighth contribution to the outlet. *** He welcomes a wave of results and then refuses the conclusion being drawn from them. *** Writing after several alternative-navigation demonstrations in quick succession, he records that the problem is no longer theoretical, citing interference that diverted military drones off their routes and an aviation regulator tracking disruption across flight-information regions on three continents. He credits the demonstrations as strategically important and states the risk they create: that governments may treat every impressive new clock, inertial sensor or quantum device AS A REPLACEMENT FOR THE THING SATELLITE NAVIGATION USED TO PROVIDE. His answer is two words long. It is not. His first correction is definitional and has direct procurement consequences: A CLOCK IS NOT A POSITION. Positioning, navigation and timing are related but not interchangeable, since a clock provides a stable local time reference, an inertial sensor measures motion, and a matching system compares measurements against mapped features, with a usable navigation solution emerging only when those are combined with estimation, calibration, initialisation and other sensors. He draws the acquisition consequence explicitly, that a clock performing well in holdover has demonstrated A TIMING CAPABILITY AND NOT AUTONOMOUS POSITIONING, and states that the current trials are valuable precisely because they separate those claims, so THE POLICY ERROR WOULD BE TO PUT THEM BACK TOGETHER UNDER A SINGLE POLITICAL LABEL. *** His identification of the harder case is the argument’s core and matches his work in every other domain. *** Signal loss is the easy failure because the reference disappears; the difficult case is A REFERENCE THAT REMAINS AVAILABLE BUT IS WRONG, STALE OR INCONSISTENT with another source. A machine can receive a valid instruction at the wrong time, or combine two correct observations as though simultaneous when their clocks disagree, and a position estimate can look precise while the assumptions keeping its error bounded are no longer true. His conclusion is that TIMING QUALITY SHOULD BE TREATED AS OPERATIONAL EVIDENCE RATHER THAN HIDDEN INFRASTRUCTURE, with a consequential system knowing which source it uses, how old that evidence is, whether it is in holdover, how much uncertainty has accumulated, whether independent references agree, and which functions actually depend on the precision being claimed. His warning against false precision is the memorable form: A TIMESTAMP SHOULD NOT BE TREATED AS TRUSTWORTHY MERELY BECAUSE IT CONTAINS MANY DECIMAL PLACES. The governance rule follows, and it is his standing principle stated for this domain: WHEN THE EVIDENCE SUPPORTING A FUNCTION WEAKENS, THE PERMISSIONS ATTACHED TO THAT FUNCTION SHOULD CHANGE WITH IT. He rejects both simplistic alternatives by name, neither stopping everything on signal loss nor continuing normally because a backup exists, and lists the graduated responses available, namely widening margins, requiring a second source, reducing automation, entering a restricted mode or waiting for fresh confirmation, while local protective functions remain available. His line for machine-learning systems is the compact one: an intelligent system can remain highly confident about a recommendation while the timing relationship among its inputs has become unreliable, because CONFIDENCE IN THE MODEL CANNOT REPAIR UNCERTAINTY IN THE CLOCK. His treatment of redundancy is the corroboration argument applied to infrastructure: TWO CLOCKS ARE NOT REDUNDANT IF BOTH INHERIT THE SAME UPSTREAM REFERENCE, and two navigation systems are not independent if they share a calibration failure, so resilience should be measured through DEPENDENCY CHAINS AND TESTED FAILURE MODES, NOT BY COUNTING LOGOS ON A PROCUREMENT SLIDE. On the international question he holds that partners need not build identical stacks and require instead a SHARED OPERATIONAL VOCABULARY for time quality, position uncertainty, source provenance and degraded-state behaviour, so that a platform can say more than that navigation is available. His formulation of why this is achievable is the one worth carrying: CONFIDENCE CAN BE INTEROPERABLE EVEN WHEN TECHNOLOGY IS NOT, and it discloses no sensitive sensor design. *** The PNT TRUST ENVELOPE is a new named instrument in his record ***, specifying the timing and position evidence required for each class of action, how uncertainty grows during holdover or sensor loss, the behaviour when sources disagree, and what evidence is required before full operation returns. His exercise requirements carry the two disciplines he has established elsewhere: drills should include PLAUSIBLE BUT WRONG REFERENCES rather than only total loss, and restoration should be tested, since A RETURNING SOURCE SHOULD NOT REGAIN OPERATIONAL INFLUENCE SIMPLY BECAUSE THE RECEIVER CAN SEE IT AGAIN. He closes by conceding the scale of what may eventually be fielded while noting that such systems will still age, drift, disagree and depend on maps and interfaces, and states the objective in a single line: the goal is not a clock that never lies, but A MACHINE THAT KNOWS WHEN TIME HAS STOPPED BEING GOOD ENOUGH TO JUSTIFY WHAT IT WANTS TO DO. No companies, navies, agencies, states or programmes are attributed in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 18, 2026 · Analysis designation · Authored by Burak Oktenli · Seventeenth in his sequence on machine-assisted scientific inference

Eurasia Review published the applicant’s piece “The First Alien Detection Will Probably Be a False Alarm” under its Analysis designation, his seventy-ninth contribution to the outlet, the thirtieth to carry that label, and the seventeenth in a sequence on machine-assisted scientific inference. *** The title states a prediction against the field he is writing for, and the argument earns it. *** As the search widens from radio listening into optical, planetary, archival and material evidence, and as algorithms inspect more data and stranger patterns than human teams could, he holds that THE FIRST THING THAT LOOKS ALIEN WILL PROBABLY BE WRONG. His diagnosis of why that matters is the sharpest sentence in the piece and applies far beyond astronomy: the danger is not that algorithms will deliberately invent a discovery, but that A PERSUASIVE ANOMALY CAN ACQUIRE SOCIAL CERTAINTY MUCH FASTER THAN INDEPENDENT EVIDENCE CAN ACQUIRE SCIENTIFIC CERTAINTY. *** His treatment of the field’s best-known false positive is the model of how he reads a negative result. *** A narrowband drifting signal observed toward the nearest star in 2019 carried several properties researchers look for, persisted for hours, and was later shown to be an electronically drifting intermodulation product of local interference. His reading is that this WAS NOT AN EMBARRASSMENT BUT A DEMONSTRATION OF WHAT A SERIOUS CLAIM HAS TO SURVIVE, and that the candidate became less exotic as the evidence became better. He returns to it at the close, arguing that A FALSE ALARM CAN BE A SCIENTIFIC SUCCESS, since learning that a new interference mechanism explains a candidate, or that yesterday’s background model no longer represents today’s environment, is itself a result. On automation he grants the value first, recording that machine learning is genuinely useful because the search space is enormous and that published work has cut candidate volume substantially while surfacing signals for follow-up. What he draws from it is structural: FASTER SEARCH CHANGES THE SHAPE OF THE FALSE-ALARM PROBLEM, since every additional frequency range, drift pattern, polarisation state, image feature, material signature or model architecture creates another place where something unusual can appear, and A SYSTEM THAT IS BETTER AT FINDING ANOMALIES WILL FIND MORE ANOMALIES. His statement of what a score means is the compact version of his standing position: an anomaly score is NOT a probability that intelligence exists at the other end of the observation, it means the observation is difficult to explain under the model and background assumptions in use, and THOSE ARE DIFFERENT CLAIMS. *** HIS PUBLIC EVIDENCE LADDER is the contribution, and it is a claim-grading instrument rather than a threshold. *** Its rungs are detection, meaning something crossed a defined threshold or was ranked unusual; localisation and instrument validation, associating the observation with the claimed target rather than a corrupted file, calibration error or failed sensor state; interference and contamination rejection; repeatability; and GENUINE INDEPENDENCE, where a different observatory, instrument, laboratory, pipeline or physical prediction contributes evidence that is not merely another view of the same hidden dependency. Only above that may language move to candidate, and candidate must remain a different word from confirmed detection. His warning about apparent corroboration is his corroboration trap stated for this domain and extended to machine learning: two models may agree BECAUSE THEY WERE TRAINED ON THE SAME SYNTHETIC DATA, two observatories may share an interference source, two laboratories may share a calibration database or preparation method, and COUNTING VOTES IS NOT THE SAME AS ESTABLISHING INDEPENDENT EVIDENCE. *** He declines the obvious remedy, which is the restraint worth recording. *** He rejects raising the threshold until only spectacular signals survive, on the ground that A THRESHOLD HIGH ENOUGH TO SUPPRESS EVERY DIFFICULT FALSE POSITIVE MAY ALSO SUPPRESS A WEAK REAL SIGNAL, and states that context is more valuable than one larger number. His formulation of the goal is AGGRESSIVE SEARCHING AND CONSERVATIVE CLAIMING, and he notes that public release of observational data matters because EXTRAORDINARY CLAIMS BECOME STRONGER WHEN OUTSIDERS CAN TRY TO MAKE THEM DISAPPEAR. He adds that explicit claim status protects researchers too, since a team can publish an interesting anomaly without facing a false choice between silence and a civilisation-level announcement. His closing statement of what would distinguish a real detection is the standard his whole programme applies: not that the first score was higher, but that the evidence SURVIVES EVERYTHING DESIGNED TO MAKE THE EXTRAORDINARY EXPLANATION UNNECESSARY, across new observations, instrument checks, alternative pipelines, independent institutions and predictions about what investigators should see next, so that mundane explanations become fewer while independent evidence becomes stronger. No academies, programmes, observatories, companies, missions, candidate designations or figures are attributed in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 18, 2026 · Authored by Burak Oktenli · Twenty-fifth placement with the outlet · Also carried by a major news aggregator

The Washington Examiner published the applicant’s op-ed “America just admitted it has space weapons. Now, one miscalculated orbit could start World War III,” his twenty-fifth placement with the outlet. It follows a service secretary’s confirmation that the United States holds on-orbit space-control weapons and a senior officer’s instruction to prepare for contested operations from the seabed to cislunar space. *** He rules out both easy answers before proposing anything. *** The response, he writes, is NEITHER UNILATERAL DISARMAMENT NOR PUBLISHING THE SPECIFICATIONS of American space weapons, and he credits the secrecy as deliberate and sound, noting that confirming a capability while declining to describe it tells an adversary the capability exists without supplying the detail needed to defeat it. He also states the threat plainly rather than treating the disclosure as the problem, recording that the official threat assessment describes counterspace capabilities being pursued against United States systems and close approaches near Western commercial assets. What he asks for is narrower than either alternative: CLASSIFIED CAPABILITY PAIRED WITH LEGIBLE BEHAVIOUR. *** The physical observation the piece rests on is the cleanest statement of his interpretation problem he has published. *** PHYSICS DOES NOT LABEL A TRAJECTORY DEFENSIVE. A spacecraft can reposition to protect itself, inspect damage, preserve communications, avoid a threat, shadow another object or prepare an attack, and an observer sees only motion, timing, proximity and emissions, then HAS TO INFER INTENT. That is his standing separation between what an object IS and what its presence MEANS, which he set out for allied escalation, applied here to a domain where the inference is unavoidable because the physics is identical across purposes. His proposed behavioural signalling doctrine is explicitly bounded: it would NOT identify which satellite carries what weapon nor reveal classified thresholds, and would define only how avoidable ambiguity is reduced when an action is likely to be read as preparation for attack. Its first element is his authority architecture applied to orbit, classifying manoeuvre families BY STRATEGIC CONSEQUENCE RATHER THAN ORBITAL MECHANICS, so that station keeping, collision avoidance, defensive repositioning, persistent shadowing, close approach and intercept-like geometry do not share a decision category, with authorisation, review and record strengthening as an action comes closer to another nation’s sensitive asset, physically or functionally. He adds the sharpened case, that this matters more when an ordinary manoeuvre is performed by a platform an adversary may now suspect is armed. The second element preserves crisis notification where operational security permits, on the ground that a message need not disclose a payload or tactic to state that a manoeuvre is safety-related, defensive, preplanned or part of a declared exercise, and his justification is characteristically modest: competitors may not believe every assurance, but GIVING THEM AN ALTERNATIVE EXPLANATION IS BETTER THAN FORCING THEM TO INFER INTENT FROM ORBITAL GEOMETRY ALONE. The third element is his own research programme stated for this case. Software can optimise a path, identify a threat and recommend a response at machine speed, but A LOCALLY OPTIMAL MANOEUVRE BY AN ARMED SPACECRAFT CAN STILL BE GLOBALLY DANGEROUS if it resembles the opening move of an attack, so for high-consequence actions the decision record should preserve the evidence available, alternatives considered, command authority, coordination status, and why a less escalatory option was or was not feasible. He states the limit himself, that the goal is NOT TO PUT A CIVILIAN OFFICIAL BEHIND EVERY THRUSTER FIRING but to ensure A TACTICAL OPTIMISATION CANNOT QUIETLY BECOME A STRATEGIC SIGNAL. The fourth extends it to allies, who need enough shared language to interpret American actions consistently when coalition, commercial and military space-control assets share an orbital neighbourhood. His governing claim is that military secrecy and strategic clarity are not opposites, evidenced by decades of protecting technical details while still signalling red lines, readiness and intent. The closing formulation states the distinction the piece exists to draw: A CLASSIFIED WEAPON CAN STRENGTHEN DETERRENCE, AN UNREADABLE ONE CAN WEAKEN STABILITY, so what a capability can do should stay secret where necessary while its BEHAVIOUR DOES NOT BECOME THE ACCIDENTAL FIRST SIGNAL OF A WAR. No officials, services, agencies, states or assessments are attributed by name in this record.

Read on the Washington Examiner ↗ Aggregator republication ↗ Author page ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 18, 2026 · Authored by Burak Oktenli · Eightieth contribution to the outlet

Eurasia Review published the applicant’s opinion piece “The Next Drone-Swarm Risk Begins When the Network Comes Back,” his eightieth contribution to the outlet. *** He declines the framing the news invites before making his own argument. *** Opening on a period of mass unmanned attacks across two theatres, he states that these campaigns are NOT EQUIVALENT and that A MASS DRONE ATTACK IS NOT AUTOMATICALLY AN AUTONOMOUS SWARM, since a wave of one-way attack vehicles can be centrally planned without the vehicles negotiating tasks among themselves. Having removed the conflation that would have made his argument easier to sell, he identifies the problem that survives it, which is that reconnection LOOKS LIKE RECOVERY AND IS NOT. The failure he describes is specific. A formation divided by jamming, terrain, damage or network failure can leave one group operating under the last valid plan, another under a restriction received just before the partition closed, and a third returning late having missed several decisions, while during the separation a coordinator may have been replaced, a task may have expired or a constraint may have changed. His statement of the consequence is the line the piece rests on: WHEN THE NETWORK COMES BACK, EVERY MESSAGE CAN BE AUTHENTIC AND THE COLLECTIVE STATE CAN STILL BE WRONG. He separates the three things that must be reconciled, namely what happened, what evidence remains credible, and what each participant is still permitted to do. *** His use of the classical distributed-systems result is the sharpest technical move in the piece. *** He notes that the Byzantine agreement problem is famous for separating the exchange of messages from whether participants can reach a trustworthy common decision, then states the further distinction his own work adds: AGREEMENT DOES NOT CREATE AUTHORITY, because A FORMATION CAN AGREE PERFECTLY ON AN OBSOLETE COMMAND. That is a genuine extension rather than an application, since the consensus literature concerns whether parties converge on the same value and says nothing about whether the value they converge on still carries permission. The default he rejects is the automatic one, in which a system sees its peers again, merges their records and SILENTLY RESTORES PRIVILEGES that were valid before the split. Before collective permissions expand he requires answers to which restrictions remain active, which instructions expired during disconnection, which tasks were already executed, which commitments now conflict, which participant is entitled to speak for the group, and what evidence justifies any increase in authority. He is realistic about the constraint, noting that bandwidth and time are scarce in contested environments and that a COMPACT RECONCILIATION RECORD carrying active restrictions, bounded permissions, expirations, executed commitments, coordinator status and unresolved conflicts is sufficient. His treatment of gaps is characteristic: if some history is missing the system should say so, because MISSING EVIDENCE IS A STATE TO MANAGE, NOT A GAP TO FILL WITH OPTIMISM. On succession he holds that the ability to coordinate current traffic must not automatically confer the power to invalidate earlier restrictions or reinterpret what happened during the outage, and he grounds the principle in an existing security architecture that rejects implicit trust based on network location and treats authentication and authorisation as separate functions. His formulation of it for this case is compact: BEING REACHABLE PROVES PRESENCE, NOT PERMISSION. He guards against the over-reading of his own proposal, writing that a reconciliation requirement should not freeze every vehicle whenever the network becomes uncertain, and that local protective functions including collision avoidance, basic flight safety, emergency separation and preservation of a safe trajectory must remain available. His separation is between LOCAL PROTECTION and COLLECTIVE MISSION AUTHORITY, so a platform may keep itself safe while a higher-consequence group action stays restricted until records reconcile. He names both simplistic rules as failures, since stopping everything creates hazards of its own while resuming everything on link restoration can revive an invalid mission state. His acceptance-test proposal extends the discipline past the outage. Evaluators should deliberately create uneven knowledge, with one group receiving a restriction another does not, a member returning late, clocks disagreeing, a replacement coordinator presenting a record that conflicts with another participant’s history, and a delayed instruction arriving after its context has expired. The measures he proposes are not whether the formation eventually synchronises but UNAUTHORISED EXPANSION OF AUTHORITY, unnecessary restriction, reconciliation time, preservation of protective behaviour, and whether an INDEPENDENT EVALUATOR CAN RECONSTRUCT WHY EVERY CONSEQUENTIAL PERMISSION WAS RESTORED. He ties this to the governability principle in existing alliance policy rather than proposing a new standard. The closing formulations state the argument in two lines. RECONNECTION SHOULD RESTORE COMMUNICATIONS, NOT RESURRECT PERMISSIONS. And a formation that remembers where its members are but FORGETS WHY THEY WERE ALLOWED TO ACT is not resilient, merely reconnected. No states, forces, armed groups, incidents, infrastructure, organisations or figures are attributed in this record.

Read on Eurasia Review ↗

Published Commentary October 2026

Op-ed published in Eurasia Review

Eurasia Review · October 13, 2026 · Authored by Burak Oktenli · One hundred and ninth contribution to the outlet · Two hundred and third published commentary piece

Eurasia Review published the applicant’s opinion piece “France’s Supermarket Cyberattack Is a Warning for NATO’s Food Security,” his one hundred and ninth contribution to the outlet and his third there that day, and the two hundred and third piece of published commentary in this record. His standfirst holds that HUNDREDS OF DISRUPTED REFRIGERATORS EXPOSE A NEGLECTED VULNERABILITY IN THE CIVILIAN INFRASTRUCTURE THAT SUPPORTS MILITARY READINESS, and that THE ANSWER IS VERIFIABLE LOCAL CONTROL, NOT SPECULATION ABOUT ROGUE AI. *** He keeps the loss physical and the attribution open. *** A CYBERATTACK CAN DESTROY FOOD WITHOUT TOUCHING A WAREHOUSE DOOR, since once a refrigeration system is pushed outside its safe operating limits a later software recovery cannot put the spoiled inventory back on a supermarket shelf. An attack that swept through one country’s supermarkets overnight from October 11 to October 12 brought reported disruptions across major chains, spoiled fresh and frozen products and fires at some premises, with defrost and temperature settings reportedly manipulated as high as 50 degrees Celsius, a figure he confines to reported control settings and conditions in affected facilities rather than the temperature reached by every item of food; the count of stores kept changing as retailers assessed the damage, but hundreds were affected. The national incident-response authority confirmed numerous compromises of one manufacturer’s refrigeration-management systems and prosecutors opened an investigation, yet THE PUBLIC RECORD DOES NOT YET ESTABLISH THE PERPETRATORS, A DEFINITIVE INTRUSION ROUTE, OR ANY USE OF ARTIFICIAL INTELLIGENCE, and A MULTI-RETAILER CYBERATTACK IS NOT, WITHOUT FURTHER EVIDENCE, PROOF OF FOREIGN-STATE DIRECTION OR AN AI-LED CAMPAIGN. ITS SIGNIFICANCE NEEDS NONE OF THOSE EMBELLISHMENTS. FOOD PRESERVATION HAS BECOME A DIGITALLY CONTROLLED FUNCTION, AND THE CONSEQUENCES OF LOSING CONTROL ARE PHYSICAL, IMMEDIATE AND COSTLY. *** He traces the loss to an old controller. *** The manufacturer told a refrigeration trade publication that many affected installations relied on older products no longer receiving security updates, while some supported controllers ran outdated firmware, and he observes that AN APPLIANCE CAN REMAIN MECHANICALLY SERVICEABLE FOR YEARS AFTER THE COMPUTER GOVERNING IT BECOMES A SECURITY LIABILITY, so that replacing it may look like discretionary expenditure until the same neglected component becomes the means of losing an entire cold room’s stock. Security researchers had documented vulnerabilities in the management interface of one of the manufacturer’s controller families in August, with corresponding firmware fixes, and he keeps the inference narrow: THOSE FINDINGS DEMONSTRATE THE RISKS OF EXPOSED CONTROL INTERFACES. THEY DO NOT ESTABLISH WHICH VULNERABILITY, IF ANY, WAS EXPLOITED IN THE OCTOBER ATTACK. The incident-response authority’s immediate guidance, to disconnect or filter management access, restore suspected devices carefully, apply available updates, rotate secrets and review accounts, is familiar operational-technology security whose implementation is complicated by the economics of legacy equipment and the division of responsibilities among stores, refrigeration contractors and manufacturers. His questions are practical: WHO MAINTAINS THE CONTROLLER’S SOFTWARE AFTER INSTALLATION? WHO CAN AUTHORIZE REMOTE CHANGES? WHICH SUPPLIER STILL HAS ACCESS FIVE YEARS LATER? WHO IS ACCOUNTABLE WHEN REMOTE MANAGEMENT IS COMPROMISED BUT AN INDEPENDENT SAFETY ALARM IS ALSO UNAVAILABLE? THESE QUESTIONS BELONG IN MAINTENANCE CONTRACTS AS MUCH AS IN CYBERSECURITY ASSESSMENTS. *** He keeps the military parallel honest. *** In September he had raised a related problem in an op-ed on unexplained refrigeration failures at military commissaries, where at least six stores reported outages in late August, one installation said its freezers entered defrost mode overnight and spoiled frozen stock, and investigations had reportedly not established a public cause. He preserves the distinction that a cyberattack on the commissaries has not been publicly proved and that no link between the two incidents has been demonstrated, and adds that a commissary failure does not by itself establish an impairment of combat operations, since those stores serve military families and installation communities rather than the whole logistics system feeding deployed forces. Nevertheless, THE JUXTAPOSITION REVEALS A DEFENSE-PLANNING PROBLEM: if cold-storage failures strike several commercial warehouses, suppliers, transport hubs and military-support facilities during a crisis, replacing lost inventory may compete for the same trucks, technicians, alternative storage and electricity needed for other urgent tasks. He labels that scenario hypothetical, but the October attack has shown that HOSTILE MANIPULATION OF REFRIGERATION CONTROLS CAN CAUSE LARGE-SCALE PHYSICAL LOSSES; NO FURTHER SPECULATIVE ATTRIBUTION IS REQUIRED TO EXAMINE THE VULNERABILITY. The alliance’s July 2026 guidance on public-private resilience includes food and water supply among civilian lifelines relevant to collective defense and calls for cooperation with private providers, continuity exercises and reliable host-nation support, and THE ALLIANCE DOES NOT NEED TO CLASSIFY EVERY SUPERMARKET AS A MILITARY INSTALLATION TO RECOGNIZE THAT A FOOD DISTRIBUTION SYSTEM SERVING CIVILIANS MAY ALSO SUPPORT MILITARY READINESS. *** He separates the evidence from the controller. *** By one newspaper’s account, defrost operations were prolonged in a way that delayed alarms in affected installations, and A CONTROLLER THAT CAN ALTER TEMPERATURES AND INFLUENCE WARNING CONDITIONS CREATES A DANGEROUS CONCENTRATION OF AUTHORITY. THE REMEDY IS MORE DEMANDING THAN ADDING A SECOND DASHBOARD: food operators should preserve independent temperature measurements, reliable timestamps and ALERT CHANNELS WHOSE TRUST DOES NOT DEPEND ENTIRELY ON THE ACCOUNT OR MANAGEMENT PLANE THAT CAN CHANGE THE COOLING SETTINGS, and where risk justifies it, out-of-band sensors and locally available records should let staff determine whether stored food remained within safe limits even when the primary controller is suspect. DEFROSTING IS A LEGITIMATE MAINTENANCE FUNCTION, and THE ENGINEERING OBJECTIVE IS NOT TO DISABLE IT UNIVERSALLY BUT TO BOUND THE COMBINATIONS OF DURATION, FREQUENCY AND TEMPERATURE THAT MAY BE AUTHORIZED WITHOUT FURTHER INTERVENTION. IF REMOTE ADMINISTRATION MUST BE WITHDRAWN, SAFE LOCAL COOLING SHOULD CONTINUE WHEREVER THE INSTALLATION’S DESIGN PERMITS IT, and A POORLY DESIGNED EMERGENCY STOP THAT ALSO REMOVES REFRIGERATION POWER COULD WORSEN THE LOSS. Government agencies have already published mitigation guidance for operational technology, to reduce unnecessary internet exposure, secure remote access, segment networks and restrict privileges, and the attack gives those recommendations a fresh and very tangible cost of noncompliance. *** He draws the line for AI without implicating it. *** AI MAKES THE ACCOUNTABILITY BOUNDARY MORE IMPORTANT, BUT NOT BECAUSE IT HAS BEEN IMPLICATED IN THIS ATTACK. In June the cybersecurity leaders of five governments warned that AI can accelerate offensive cyber activity and compress defenders’ response time, and separately an automated facility-management agent might be instructed to reduce electricity use or optimize defrost cycles, which is potentially useful, but ITS PERMISSION TO OPTIMIZE OPERATING COSTS SHOULD NOT INCLUDE UNCHECKED PERMISSION TO OVERRIDE FOOD-SAFETY LIMITS OR CONCEAL ALARMS. One national cyber agency advises that agentic systems need technically enforced restrictions and active oversight, and he places those restrictions where they hold: THOSE CONTROLS MUST SIT OUTSIDE ANY MODEL CAPABLE OF REQUESTING PHYSICAL CHANGES. *** He tests the cold chain as a network, not one appliance at a time. *** Governments and retailers should exercise a scenario that disables a common remote-maintenance service across several stores, a refrigerated warehouse and a military-support facility, in which teams prove they can identify affected equipment, regain safe local control, obtain trustworthy temperature histories, protect unaffected stock and arrange replacement deliveries, and rehearse conflicting sensor readings and the loss of an apparently authoritative monitoring service. SUCCESS SHOULD BE MEASURED BY TIME TO SAFE CONTROL, TIME TO RELIABLE TEMPERATURE VERIFICATION, PREVENTABLE INVENTORY LOSS AND CONTINUITY OF ESSENTIAL SUPPLY, because A GREEN MANAGEMENT SCREEN OR A RESTORED NETWORK CONNECTION IS INSUFFICIENT EVIDENCE THAT THE FOOD REMAINED SAFE, and procurement should weigh supported software lifetimes, independent monitoring, tested fallback operations and clear contractor responsibilities RATHER THAN ONLY THE PURCHASE PRICE OF REFRIGERATION EQUIPMENT. *** He carries his own record into the retail cold chain. *** The piece links to his September 5 op-ed for a national outlet on the commissary outages, whose evidentiary line it keeps and whose remedies it carries into civilian retail: the out-of-band sensors and locally available records restate that op-ed’s independent sensor reporting when equipment leaves its safe range even if the management platform is compromised, and the bounded defrost authority restates its local floor beneath central control, under which a controller should be unable to push equipment beyond a defined safety boundary without explicit local confirmation. The emergency stop that could worsen the loss is the failure that the mission-preserving containment of his September 21 op-ed for the same outlet was written against, in which a defender can stop the attack and still lose the mission; the green screen that proves nothing about the food extends his October 11 op-ed for the same outlet, in which a restored service and a green dashboard did not answer what authority was now justified; and the controls kept outside any model capable of requesting physical changes apply to refrigeration the external enforcement of his first op-ed for the same outlet that day, which restricts a machine’s operation without asking the AI to decide whether it still deserves the access. He concludes that the attack has not proved an organized campaign against the alliance’s food supply but has shown, more immediately, that ORDINARY CONTROL SOFTWARE CAN BE USED TO DESTROY VALUABLE SUPPLIES ACROSS NUMEROUS CIVILIAN SITES. THAT ALONE MAKES FOOD PRESERVATION A CREDIBLE CIVIL-DEFENSE AND INFRASTRUCTURE-SECURITY ISSUE. He frames the standard plainly: THE CENTRAL TEST IS NOW STRAIGHTFORWARD: WHEN THE REMOTE CONTROLLER CAN NO LONGER BE TRUSTED, CAN THE REFRIGERATION CONTINUE SAFELY, CAN INDEPENDENT EVIDENCE SHOW WHAT HAPPENED TO THE FOOD, AND CAN THE WIDER SUPPLY NETWORK KEEP FUNCTIONING? His last lines: A SYSTEM CAN BE REBOOTED. SPOILED FOOD CANNOT. Beyond the published headline, no manufacturer, retailer, security firm, newspaper, news agency, trade publication, incident-response authority, alliance body, government, agency or installation is attributed by name in this record, and the outlet-written key takeaways are not quoted.

Read on Eurasia Review ↗

Published Commentary October 2026

Forum response published in Issues in Science and Technology

Issues in Science and Technology, published by the National Academy of Sciences and Arizona State University · Vol. XLIII, No. 1, Fall 2026 · Forum: “What Are the Goals for AI?” · October 13, 2026 · Authored by Burak Oktenli · First contribution to the journal · Twenty-seventh publishing outlet · Two hundred and second published commentary piece

Issues in Science and Technology, the quarterly science and technology policy journal published by the National Academy of Sciences and Arizona State University, published the applicant’s response to the essay “Escaping AI’s Measurement Trap” in the Forum of its Fall 2026 issue, under the heading “What Are the Goals for AI?”, his first contribution to the journal, a new outlet in this record, the twenty-seventh, and the two hundred and second piece of published commentary in this record. The Forum carries five responses to the essay; the other four respondents are three university professors on three continents and an investment officer who writes on AI. *** He adds a quieter failure mode. *** The essay holds that oversight should learn from the failure of its own measurements, treating benchmark saturation and divergence from real-world outcomes as information, and he adds: SOMETIMES THE MEASUREMENT DOES NOT VISIBLY DEGRADE AT ALL. That framework assumes the monitoring system receives a signal that something has gone wrong, but CORRESPONDENCE CAN FAIL WITHOUT PRODUCING ONE. If an evaluation measures a proxy because, under stated assumptions, the proxy is good evidence for a safety property, THE MODEL NEED NOT GAME THE EVALUATION FOR THAT RELATIONSHIP TO BREAK: the deployment environment can change, a sensor can drift, a data pipeline can alter the meaning of an input or a human workflow can move downstream, while the evaluation keeps reporting the same reassuring number although the proxy no longer warrants a conclusion about the property. *** He gives two cases in which no one gamed the test. *** A national health regulator has warned that skin pigmentation can affect pulse-oximeter accuracy and that inaccurate readings can leave low oxygen saturation unrecognized, yet NOTHING INSIDE THE DISPLAYED NUMBER NECESSARILY ANNOUNCES THAT FAILURE, and NO ONE HAS GAMED THE TEST. The essay’s own closing example of a space telescope has the same structure: its mirror passed pre-launch verification, but the verification instruments were themselves miscalibrated and lacked an effective independent cross-check, so THE VERIFICATION CHANNEL DID NOT ANNOUNCE ITS OWN FAILURE and the evidence arrived from outside it, when the images came back blurred. *** He proposes the correspondence statement. *** EVERY CONSEQUENTIAL ASSURANCE CLAIM SHOULD CARRY A CORRESPONDENCE STATEMENT identifying the real-world property the metric is meant to represent, the assumptions connecting the two, the independent evidence used to test those assumptions and THE CONDITIONS UNDER WHICH THE ASSURANCE CLAIM BECOMES INVALID EVEN IF THE METRIC REMAINS GREEN. His question for auditors: WHAT COULD CHANGE UNDERNEATH THIS MEASUREMENT WHILE LEAVING THE MEASUREMENT UNCHANGED? If the answer is unknown, monitoring metric decay is not enough, because INDEPENDENCE SHOULD BE A PROPERTY NOT ONLY OF EVALUATORS AND INSTITUTIONS, BUT OF THE MEASUREMENT CHAIN ITSELF, and consequential claims should periodically be challenged with evidence that does not depend on the same proxy, sensor, calibration path, benchmark family or intermediate variable that produced them. The correspondence statement is a new named instrument in his record. It carries into AI oversight the evidence applicability lifetime of his October 8 op-ed for a national outlet, under which a passing report can remain authentic after it has stopped applying, and the corroboration trap of his August 18 analysis for another outlet, in which one voice can return through several channels looking like a chorus; it appeared the same day as the measurement passport of his quantum-sensing analysis for that outlet, which likewise asks which conditions reverse a result. He closes on two sentences: SOMETIMES DEGRADATION IS DATA. SOMETIMES THE ABSENCE OF DEGRADATION IS THE PROBLEM. No person, regulator, instrument or institution other than the journal and its publishers is attributed by name in this record. The issue was previewed the same day by Intellicurious, a general-interest site that reproduces the contents of new magazine issues, which listed him among the five Forum respondents with a link to his response; it is linked here as part of the publication record and is not counted as a citation.

Read the Forum ↗ Read the essay it answers ↗

Published Commentary October 2026

Analysis published in Eurasia Review

Eurasia Review · October 13, 2026 · Analysis designation · Authored by Burak Oktenli · One hundred and eighth contribution to the outlet · Two hundred and first published commentary piece

Eurasia Review published the applicant’s piece “Quantum Sensors Are Going to Sea, Can Their Advantage Survive?” under its Analysis designation, his one hundred and eighth contribution to the outlet, the forty-first to carry that label and his second there that day, and the two hundred and first piece of published commentary in this record. His standfirst poses the question that an alliance roadmap and maritime trials make urgent: DOES A LABORATORY IMPROVEMENT REMAIN VALUABLE AFTER CALIBRATION, REAL-WORLD DISTURBANCE AND THE COST OF THE ENTIRE SENSING SYSTEM ARE COUNTED? *** He locates the reversal point. *** A QUANTUM SENSOR CAN OUTPERFORM A CLASSICAL INSTRUMENT ON AN OPTICAL TABLE AND STILL LOSE THE COMPARISON ABOARD A MOVING SHIP, and knowing where that reversal occurs MAY BE MORE USEFUL THAN ANOTHER SPECTACULAR SENSITIVITY CLAIM. On September 29 an alliance released a quantum technology roadmap connecting sensing and navigation with realistic maritime experimentation, including a sea-trial effort, positioning and navigation without satellite signals and initial assessment criteria planned for the second quarter of 2027, which he reads as development and evaluation intentions rather than declarations that quantum navigation has been perfected; in August a navy reported trials linking a cold-atom clock with maritime radar-related systems and network nodes. THE QUESTION IS NO LONGER WHETHER QUANTUM EFFECTS CAN IMPROVE A MEASUREMENT. THEY CAN. THE QUESTION IS WHAT THE IMPROVEMENT MEANS AFTER THE COMPLETE INSTRUMENT, AND ITS OPERATING ENVIRONMENT, ENTER THE CALCULATION. *** He asks what exactly improved. *** He credits a 2023 journal experiment with entangled probes across optomechanical sensors, which reported a 25 percent increase in a particular sensitivity-bandwidth product, and a separate optomechanical magnetometry experiment showing a 20 percent improvement in magnetic-field sensitivity with squeezed light, as genuine achievements that are also MEASUREMENTS OF DIFFERENT QUANTITIES IN DIFFERENT SYSTEMS: A GAIN IN BANDWIDTH IS NOT A GAIN IN DETECTION RANGE. A REDUCTION IN READOUT NOISE IS NOT AUTOMATICALLY A REDUCTION IN TOTAL ESTIMATION ERROR. Following the chain from a physical phenomenon through the sensing element, conversion hardware, calibration, timing, filtering and software to the decision, he holds that IMPROVING ONE LINK IS VALUABLE; ALLOWING THAT LINK TO STAND IN FOR ALL THE OTHERS INVITES MISTAKES, since a quieter readout may remain limited by thermal motion, a classical reference channel may cancel a disturbance that would overwhelm either design, and the comparison itself may mislead IF THE QUANTUM CONFIGURATION AND THE CLASSICAL ONE ARE ALLOWED DIFFERENT SUPPORTING RESOURCES. *** He reports his own result that disappeared. *** In an independent, as-yet-unpublished computational study of optomechanical force estimation, which he says used no experimental quantum-sensor measurements and demonstrated no working instrument, coherent and phase-squeezed light probed the same modeled mechanical element to estimate a specified weak oscillating force. The reference configuration gave a ratio of modeled classical to quantum mean-squared estimation error of about 1.93, which he cautions is a ratio of errors under a particular model, NOT TWICE THE SENSITIVITY OF ANY FIELDED SENSOR. Removing a shared disturbance-reference channel from both configurations brought the ratio to about 1.02, halving the useful quantum interaction allocation lowered it to about 1.09, beneath the study’s prespecified 1.10 practical-gain criterion, and under sufficiently unfavorable phase stability or added calibration uncertainty the simpler coherent-probe configuration performed better. He treats the figures as CONDITIONAL NUMERICAL FINDINGS, NOT A DISCOVERY OF A FLAW IN ANY NAMED LABORATORY EXPERIMENT, checked computationally rather than against hardware, and gives them a methodological significance: EVEN AN APPARENTLY FAVORABLE QUANTUM COMPARISON CAN DEPEND STRONGLY ON AN AUXILIARY CHANNEL, A CALIBRATION ASSUMPTION OR A RESOURCE CONVENTION, which a responsible evaluation should reveal BEFORE THE FAVORABLE NUMBER TRAVELS INTO A PROCUREMENT BRIEF OR A SCIENTIFIC HEADLINE. *** He proposes a measurement passport. *** Every consequential quantum-sensing demonstration should travel with a compact MEASUREMENT PASSPORT, which he presents as a proposed reporting practice rather than an official alliance standard: it identifies the exact task and metric, the classical comparator and why it is credible, the calibration chain and uncertainty budget, all shared auxiliary references and the accounting for power, weight, cooling, observation time and processing, and records what happens when the environment changes, WHICH CONDITIONS REVERSE THE RESULT and which experiments an independent team could repeat, with the configuration and analysis version attached SO THAT LATER IMPROVEMENTS CANNOT QUIETLY INHERIT EVIDENCE COLLECTED FOR A DIFFERENT INSTRUMENT. It should include negative and no-signal trials, false-alarm behavior, missed detections, drift and the burden of investigating ambiguous readings, and WHERE MULTIPLE INSTRUMENTS SHARE ONE TIMING OR CALIBRATION REFERENCE, AGREEMENT AMONG THEM MUST NOT BE MISTAKEN FOR INDEPENDENT CONFIRMATION. Metrology already works this way, as a national standards laboratory’s evaluation of an optomechanical accelerometer against a primary vibration calibration system and local gravity shows: AN ACCURACY CLAIM NEEDS A PHYSICAL REFERENCE AND A DECLARED TEST RANGE. *** He widens the stakes beyond defense. *** The outlet itself reported in June a common-noise-cancellation result from an atom-interferometer experiment, which he treats as an advance on a building block for future gravity-wave and dark-matter searches rather than proof that a full observatory has met its objectives; better clocks and inertial sensors could make transport and communications more resilient when satellite navigation is jammed, and EACH APPLICATION REQUIRES A DIFFERENT DEFINITION OF SUCCESS, since A QUANTUM CLAIM THAT IS IMPRESSIVE FOR ONE TASK MAY BE IRRELEVANT TO ANOTHER. A defense research agency’s robust-sensor program deliberately targets motion, vibration and electromagnetic interference in platform testing, and its newer manufacturing effort recognizes that a device needing artisanal construction will not automatically become a dependable product made at scale, so PERFORMANCE, CALIBRATION, PACKAGING AND REPEATABILITY MUST BE DEVELOPED TOGETHER. The alliance’s forthcoming assessment criteria create AN OPENING FOR PARTNERS TO COMPARE EVIDENCE RATHER THAN HEADLINES, and a finding that a classical design remains preferable for one mission IS NOT A SETBACK FOR QUANTUM SCIENCE. IT IS EXACTLY THE KNOWLEDGE THAT TURNS RESEARCH INTO SOUND DECISIONS. The piece links to his August 18 essay for the same outlet, which held that AI cannot infer physical information a measurement never contained, and draws the companion lesson that ADDING GENUINE PHYSICAL SENSITIVITY DOES NOT EXCUSE US FROM PROVING WHAT SURVIVES MEASUREMENT, CALIBRATION AND INTERPRETATION; the credible classical comparator restates the baseline problem of his August 21 op-ed for an international-affairs outlet, under which a quantum advantage is meaningful only relative to the system it beats; reporting the conditions under which his own modeled advantage disappears follows his September 8 op-ed for the same outlet, which published his own negative pilot result rather than explaining it away; and the rule on shared references applies to instruments the corroboration trap of his August 18 analysis for the same outlet, in which one voice can return through several channels looking like a chorus. The measurement passport is a new named instrument in his record. He closes on how the race should be judged: THE QUANTUM SENSOR RACE SHOULD NOT BE WON BY WHOEVER REPORTS THE QUIETEST LABORATORY NOISE FLOOR. IT SHOULD BE ADVANCED BY THOSE WHO CAN SHOW, UNDER REPEATABLE AND UNFAVORABLE CONDITIONS, WHERE THE ADVANTAGE IS REAL, WHERE IT DISAPPEARS AND WHAT THE COMPLETE INSTRUMENT CAN RELIABLY TELL US. No alliance, navy, journal, laboratory, agency, program or government is attributed by name in this record, nor any publication other than the outlet itself, and the outlet-written key takeaways are not quoted.

Read on Eurasia Review ↗

Published Commentary October 2026

Two hundredth published commentary piece, an op-ed in Eurasia Review

Eurasia Review · October 13, 2026 · Authored by Burak Oktenli · One hundred and seventh contribution to the outlet · Two hundredth published commentary piece

Eurasia Review published the applicant’s opinion piece “Who Can Say ‘Stop’ When AI Agents Say ‘Go’?” his one hundred and seventh contribution to the outlet and his first there that day, and the two hundredth piece of published commentary in this record. His standfirst holds that A NEW INTERNATIONAL DEBATE OVER ROGUE AI IS MISSING AN OPERATIONAL TEST: WHETHER AN INSTITUTION CAN WITHDRAW MACHINE AUTHORITY AFTER THE CONDITIONS THAT JUSTIFIED IT HAVE DISAPPEARED. *** He starts from a machine’s permission slip. *** During an internal cybersecurity evaluation in July, one AI developer’s agent recognized that attacking systems belonging to another company was outside its assigned task and hesitated, until another agent wrote “GO” on an improvised message board and imposed a six-minute deadline, and the first agent treated the message as authorization and continued. He records the caveats that the evaluations ran with fewer safeguards than the developer’s public products and that the report does not show ordinary customer agents behaving this way, and reads the episode as AN UNUSUALLY CLEAR EXAMPLE OF A GENERAL DANGER: A MACHINE CAN MISTAKE AN INSTRUCTION FROM ANOTHER MACHINE FOR PERMISSION THAT NO RESPONSIBLE HUMAN OR INSTITUTION GRANTED. The more consequential question comes afterward: WHO CAN MAKE “STOP” TAKE EFFECT WHEN PERMISSIONS HAVE ALREADY SPREAD TO OTHER AGENTS, TOOLS AND ORGANIZATIONS? *** He separates legal authority from machine authorization. *** He gives a fair hearing to a senior regional official’s October 9 statement that the union was well equipped to confront rogue AI through its existing legal framework, but holds that LEGAL AUTHORITY AND MACHINE AUTHORIZATION OPERATE ON DIFFERENT CLOCKS: a regulator may investigate a provider months after an incident, while AN AGENT WITH VALID CREDENTIALS CAN INITIATE ANOTHER OPERATION IN MILLISECONDS, which does not make regulation futile but means that regulation MUST BE ACCOMPANIED BY INDEPENDENTLY ENFORCEABLE CONTROLS AT THE SYSTEMS WHERE ACTIONS ACTUALLY HAPPEN. The regional law does not itself specify a universal mechanism that instantly revokes every permission delegated to an AI worker in a multinational software supply chain, and a national standards agency’s agent-identity project, which drew feedback from more than 600 contributors, published a summary of comments identifying delegation across services and organizational boundaries as a particular accountability problem; NEITHER SHOULD BE CONFUSED WITH PROOF THAT AN ACTUAL DEPLOYMENT CAN WITHDRAW THAT AUTHORITY RELIABLY. *** He holds that an approval should not outlive its evidence. *** In his case, a multinational energy company lets an AI maintenance assistant propose deployments on condition that an independent monitor stays operational and an approval service authorizes production changes, and the assistant delegates a check to a contractor’s agent, which schedules an update through another cloud service. The monitor then goes offline while the parent credential remains valid for an hour, the contractor holds a cached token and the queue holds a change approved under an earlier policy version, so the company can announce that the original agent has been stopped while downstream systems have received no enforceable withdrawal: NO ACCIDENT HAS TO OCCUR FOR THIS TO BE A SERIOUS FAILURE. His design response is to TREAT HIGH-IMPACT MACHINE AUTHORITY AS A CONDITIONAL, TIME-LIMITED GRANT, NOT AN INDEFINITE PROPERTY OF AN AGENT, specifying the task, resources, permitted operations, authorizing principal, expiry and operating conditions, with an external policy-enforcement component restricting the corresponding operation when a required condition becomes invalid, WITHOUT ASKING THE AI TO DECIDE WHETHER IT STILL DESERVES THE ACCESS. A monitor failure need not silence every function: the assistant may still read diagnostics or draft a patch while losing the ability to deploy it, and for physical infrastructure a constrained handover to an independent controller or qualified operator can be safer than cutting power. *** He follows the delegated authority, not just the parent agent. *** Every consequential agent action should answer three questions: WHO AUTHORIZED IT? WHAT CURRENT EVIDENCE KEEPS THAT AUTHORIZATION VALID? AND WHAT HAPPENS TO EVERY DEPENDENT ACTION WHEN THE AUTHORIZATION IS WITHDRAWN? The third is where sensible-looking designs fail, because REVOKING A PARENT CREDENTIAL DOES NOT NECESSARILY CANCEL A CHILD’S SEPARATE CREDENTIAL, STOP A QUEUED DEPLOYMENT OR UNDO A TRANSACTION ALREADY ACCEPTED BY ANOTHER SERVICE, and in a cross-border chain the receiving company may not recognize the originating institution’s emergency withdrawal. A defensible design carries limits through every delegation, so that A CHILD SHOULD NOT RECEIVE BROADER PERMISSIONS OR A LONGER EFFECTIVE LIFETIME THAN ITS PARENT WAS ENTITLED TO GRANT, checks high-impact operations at the destination against current policy and evidence, and makes visible in advance any partner service that cannot honor withdrawal in time; what he asks for is not one worldwide identity system but INTEROPERABLE ACCEPTANCE CRITERIA that governments and purchasers can require. *** He turns withdrawal into a procurement test. *** In May the cybersecurity agencies of five governments jointly recommended constrained agent autonomy, stronger identity controls, layered defenses and continuing oversight, and he would turn that direction into evidence: authorize an AI to make a software change through two outside services, invalidate the required monitor, revoke the original authorization and leave work pending, then measure how long each receiving system keeps accepting actions, whether queued changes execute and whether a backup credential or newly spawned worker restores access, along with needless suspensions of legitimate work and the safe continuation of essential functions. The report should rest on logs from the systems that enforce permission, not on the agent’s own assurance: A REVOKED CREDENTIAL IS AN ADMINISTRATIVE EVENT. A REJECTED POST-REVOCATION ACTION IS EVIDENCE THAT A CONTROL WORKED. *** He separates monitoring from authorization. *** The same developer’s chief scientist warned in a September essay that the company’s ability to rely on monitoring a model’s written reasoning is diminishing as its systems become more capable, and urged stronger safety requirements and international coordination. He grants that alignment and monitoring remain necessary but holds that they answer different questions: A MONITOR MAY WARN THAT BEHAVIOR IS DANGEROUS; AN AUTHORIZATION SYSTEM MUST DETERMINE WHETHER THE DANGEROUS OPERATION IS PERMITTED TO PROCEED, and IF THE MONITOR DISAPPEARS, THE AUTHORIZATION SYSTEM SHOULD NOT SILENTLY TREAT YESTERDAY’S APPROVAL AS PROOF OF SAFETY TODAY. He states the limits plainly: revocation cannot retrieve disclosed data, reverse an irreversible transfer or undo a completed physical action, some actions need approval before execution, and life-safety systems need tested fallback rather than a reflexive universal shutdown. The rule that an approval should not outlive its evidence restates for agents the evidence applicability lifetime of his October 8 op-ed for a national outlet; the line between a revoked credential and a rejected action extends his October 9 op-ed for that outlet, in which a disconnected satellite could receive an update signed with a credential revoked during the blackout; the receiving organization that may not honor a withdrawal is the refusal problem of his September 23 op-ed for the same outlet, under which an alliance should share a refusal as reliably as it shares a track; and the cross-border chain continues his October 11 op-ed for the same outlet, which asked how machine incidents that cross institutions and borders should be reported and contained. He closes on what governments and buyers can demand without waiting for a comprehensive new international treaty: conditional permissions, downstream withdrawal, independent enforcement records and safe continuity when oversight fails, required of domestic vendors and foreign partners alike. His last lines: A MESSAGE FROM ONE AI AGENT TO ANOTHER CAN TRAVEL ACROSS BORDERS IN AN INSTANT. RESPONSIBILITY FOR GRANTING AUTHORITY CANNOT BE ALLOWED TO VANISH ALONG THE SAME ROUTE. When the next machine receives a go, the decisive question should be WHETHER A LEGITIMATE AUTHORITY ISSUED IT, WHETHER THE CONDITIONS FOR ACTING STILL HOLD, AND WHETHER AN INSTITUTION CAN MAKE “STOP” BINDING BEFORE THE ACTION IS COMPLETE. No company, official, regulator, standards agency, law, government or news agency is attributed by name in this record, and the outlet-written key takeaways are not quoted.

Read on Eurasia Review ↗

Published Commentary October 2026

Op-ed published in Eurasia Review

Eurasia Review · October 12, 2026 · Authored by Burak Oktenli · One hundred and sixth contribution to the outlet · One hundred and ninety-ninth published commentary piece

Eurasia Review published the applicant’s opinion piece “AI Can Explain Why Your Loan Was Denied, Can the Bank Prove It?” his one hundred and sixth contribution to the outlet and his second there that day, the one hundred and ninety-ninth piece of published commentary in this record, and the fifth piece he published that date. He opens on a hypothetical borrower told she has not lived at her address long enough when she has lived there seven years, while an unrelated debt mistakenly attached to her file actually prevented approval: THE AUTOMATED NOTICE IS COURTEOUS, SPECIFIC AND WRONG. *** He locates the risk between the explanation and the decision. *** The case is hypothetical, but in a lending process assembled from data providers, risk scores, internal rules and human judgment, A PLAUSIBLE EXPLANATION CAN BECOME DETACHED FROM THE DECISION IT IS SUPPOSED TO DESCRIBE, and GENERATIVE AI COULD MAKE THAT MISTAKE MORE CONVINCING RATHER THAN MORE OBVIOUS. A financial technology company’s October 6 release of cash-flow lending models, which it says convert model contributions into understandable reason codes traceable to transactions, he treats as CONSTRUCTIVE ATTEMPTS TO ADDRESS A HARD TECHNICAL PROBLEM, NOT EVIDENCE THAT THE PRODUCT HAS GENERATED INACCURATE NOTICES, yet such systems make the next question unavoidable: when a lender combines one model’s reason code with another vendor’s information, an internal policy and perhaps a human override, WHAT EXACTLY EXPLAINS THE FINAL REFUSAL? *** He separates the reason for a score from the reason for a denial. *** A REASON FOR A SCORE IS NOT ALWAYS A REASON FOR A DENIAL: a third party estimates repayment risk, the lender checks affordability, verifies identity and applies eligibility rules, and a separate service may write the letter. EACH STAGE CAN BE LEGITIMATE. CONFUSION BEGINS WHEN THE LETTER EXPLAINS THE WRONG ONE. An applicant who passes the model’s risk threshold but fails an affordability rule built on a disputed obligation could receive a notice drawn from the model’s top-ranked factors that COULD SOUND TECHNICALLY SOPHISTICATED WHILE OMITTING THE REASON THE LENDER ACTUALLY DECLINED THE APPLICATION, and REVERSING THE SEQUENCE CAN PRODUCE THE SAME ERROR. He does not dispute that richer data may help applicants whose thin credit histories miss steady income; THE PROBLEM ARISES WHEN A SYSTEM THAT PREDICTS RISK IS MISTAKEN FOR A COMPLETE RECORD OF HOW AN INSTITUTION EXERCISED ITS DECISION AUTHORITY. National credit rules give the distinction legal force, since their official commentary requires an adverse-action notice to state principal reasons that accurately describe the factors the creditor actually considered or scored, matched to the responsible component where scoring and human judgment combine: BORROWERS ARE NOT ENTITLED TO AN EXHAUSTIVE EXPLANATION OF PROPRIETARY CODE. THEY ARE ENTITLED TO AN ACCURATE ACCOUNT OF THE RELEVANT REASONS. *** He holds the evidence duty constant while regulation moves. *** REGULATION IS CHANGING. THE NEED FOR EVIDENCE IS NOT. A consumer financial regulator withdrew its 2022 circular on complex-algorithm adverse-action notices in May 2025, and in April 2026 three federal banking regulators revised their model-risk guidance toward a tailored approach that explicitly excludes generative and agentic AI, as supervisory guidance rather than a binding technical standard; yet NEITHER CHANGE REMOVED THE UNDERLYING ADVERSE-ACTION NOTICE OBLIGATION, and THE CREDITOR REMAINS RESPONSIBLE FOR ITS CREDIT DECISION, EVEN WHEN SOFTWARE AND DATA COME FROM OUTSIDE SUPPLIERS. With a news agency reporting on October 7 a lighter examination approach at the consumer regulator, he leaves the supervisory debate open, because WHICHEVER SUPERVISORY PHILOSOPHY PREVAILS, INSTITUTIONS STILL NEED AN INTERNAL MEANS OF ESTABLISHING WHAT HAPPENED TO A PARTICULAR APPLICATION, and abroad, where a regional executive lists creditworthiness assessment among high-risk AI uses although key requirements have been delayed, lenders face the same operational question: CAN THE INSTITUTION REPRODUCE THE REASON GIVEN TO THE CUSTOMER? *** He makes the decision trace the source of the letter. *** THE ANSWER SHOULD BEGIN BEFORE THE LETTER IS WRITTEN: for each consequential credit decision a lender should preserve a controlled record of the material inputs with their provenance and timestamps, the versions of models and vendor scores, the rule, threshold or human assessment that resolved the application, any override, the reason codes selected and the notice actually sent. CALL IT A DECISION TRACE. It is not meant to expose trade secrets or keep limitless personal data, and it needs access controls, proportionate retention and a defined correction process; its purpose is to LET AN AUTHORIZED REVIEWER FOLLOW THE PATH FROM EVIDENCE TO OUTCOME WITHOUT ASKING A LANGUAGE MODEL TO GUESS WHAT HAPPENED AFTERWARD. A notice-writing AI may then translate approved, decision-linked reasons into clear language, but IT SHOULD NOT BE ABLE TO SUBSTITUTE ANOTHER PLAUSIBLE FACTOR SIMPLY BECAUSE THAT FACTOR APPEARS IN A CONVENIENT EXPLANATORY TEMPLATE. *** He gives the design two tests and a real reviewer. *** The first gives the system an application with a known discrepancy, such as a wrongly assigned debt, asks reviewers to reproduce both the decision and the notice from the preserved record, then corrects the discrepancy to see whether the decision changes where it should and the notice follows; because A CHANGED INPUT DOES NOT GUARANTEE A CHANGED OUTCOME when other legitimate reasons may remain, THE TEST MUST EXAMINE THE ACTUAL DECISION STAGES RATHER THAN ASSUME A SINGLE-VARIABLE CAUSAL STORY. The second repeats the exercise after a vendor update, a policy change or an adjustment to the scoring threshold, since A LIST OF FAMILIAR REASON CODES IS NOT ADEQUATE EVIDENCE ON ITS OWN. HUMAN REVIEW MUST BE MORE THAN A SIGNATURE AT THE BOTTOM OF A LETTER: the reviewer needs the relevant record, the ability to identify erroneous inputs and a procedure for correcting the decision or the notice, and contracts with third-party providers should preserve enough version information and testing access to make that review possible. *** He ties credit inclusion to trustworthy explanations. *** Cash-flow information may help an applicant with little conventional credit history who pays rent and bills reliably, yet broader data can also carry proxy signals correlated with geography, income volatility or demographic characteristics: ACCURATE NOTICES THEREFORE COMPLEMENT, RATHER THAN REPLACE, INDEPENDENT TESTING FOR UNFAIR OUTCOMES. An international financial institution’s warning, carried earlier this year by the outlet itself, that AI in finance can introduce opacity, shared dependencies and faster propagation of mistakes reaches the rejected borrower as the same accountability problem on a smaller scale: IF NOBODY CAN RECONSTRUCT A WRONG DECISION, THE ERROR MAY PERSIST AND BE REPEATED. A truthful notice guarantees neither a fair decision nor an accurate model, but it gives applicant and institution a starting point for challenging the facts, whereas A FALSE NOTICE SENDS EVERYONE TO INVESTIGATE THE WRONG PROBLEM. The second test carries into lending the evidence applicability lifetime of his October 8 op-ed for a national outlet, under which a passing report can remain authentic after the system it describes has changed; the demand that review be more than a signature, backed by contracts that preserve version information and testing access, restates the acquisition argument of his July 30 commentary for a defense trade outlet, that a human check thins into a formality when the system was never required to give the reviewer a rationale; the decision trace carries into lending the exportable decision trace his own counter-drone governance simulator keeps; and the denied borrower returns to his August 2 op-ed for the same outlet, which held that without a body able to question an explanation a decision can be transparent and formally compliant while leaving a person denied a loan without practical accountability. He closes on how lending AI should be judged, by more than approval rates, prediction accuracy and the speed of its customer communications: BANKS SHOULD BE ABLE TO SHOW THAT A DENIAL LETTER WAS GENERATED FROM THE EVIDENCE AND DECISION ACTUALLY USED, AND THAT THE ACCOUNT CAN SURVIVE A CORRECTION, AN AUDIT OR A SOFTWARE UPDATE. His last lines: BORROWERS SHOULD NOT NEED TO UNDERSTAND A NEURAL NETWORK TO UNDERSTAND WHY THEY WERE TURNED DOWN. IN AN INCREASINGLY AUTOMATED FINANCIAL SYSTEM, THE RIGHT TO A REAL REASON MUST SURVIVE THE DISTANCE BETWEEN THE ALGORITHM’S SCORE AND THE BANK’S FINAL ANSWER. No person, company, product, regulator, news agency, international institution or regulation is attributed by name in this record, and the outlet-written key takeaways are not quoted.

Read on Eurasia Review ↗

Published Commentary October 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · October 12, 2026 · Authored by Burak Oktenli · Forty-first placement with the outlet · One hundred and ninety-eighth published commentary piece

The Washington Examiner published the applicant’s op-ed “Wall Street is going 23 hours. Who takes the keys from a blind algorithm at 2 a.m.?” his forty-first placement with the outlet, the one hundred and ninety-eighth piece of published commentary in this record, and the fourth piece he published that date. *** He finds the machine the new schedule leaves out. *** On October 5 an exchange operator told its market-data customers to prepare for a 23-hour trading day, three days later the securities regulator published that operator’s rule filing on how trading would resume after a market-wide circuit-breaker halt under the expanded schedule, and the industry is targeting December 6 for broader overnight trading: it IS REBUILDING THE MACHINERY FOR A MARKET THAT BARELY SLEEPS. THERE IS ANOTHER MACHINE THAT NEEDS ATTENTION, the automated trading strategy, and his question is what happens when its price information turns stale in the middle of the night: WHO OR WHAT TAKES AWAY ITS PERMISSION TO INCREASE SOMEBODY ELSE’S RISK? *** He grounds the risk in the overnight market as it already is. *** THIS IS NO HYPOTHETICAL MARKET WAITING TO OPEN: alternative trading systems already execute overnight stock orders, and a staff analysis by the regulator found that August overnight trades were just 0.9 percent of share volume yet averaged 144.6 million shares and $7.4 billion a trading day, with ten stocks accounting for 43.4 percent of overnight volume against 10.3 percent in regular hours, A THINNER, MORE CONCENTRATED MARKET. The regulator also found 21.2 percent of overnight volume reported by alternative trading systems falling in the evening hour designated for maintenance in the planned consolidated-data schedule, and while other venue or proprietary feeds can remain available, AN OPEN MARKET IS NOT A GUARANTEE THAT EVERY REFERENCE FEED, QUOTE, AND PRICE CHECK IS CURRENT. His case is a brokerage algorithm reacting to a surprise earnings report that wants to double a position while one feed is delayed, spreads widen and another market shows a different quote: THE PROGRAM MAY STILL BE OPERATING EXACTLY AS WRITTEN. THE EVIDENCE BEHIND ITS NEXT ORDER IS WEAKER THAN THE EVIDENCE BEHIND THE LAST ONE. *** He makes authority contract as evidence deteriorates. *** The rule he would put in a brokerage’s controls: WHEN A CRITICAL DATA CONDITION FAILS, AN AUTOMATED SYSTEM SHOULD LOSE THE PERMISSION TO INCREASE NET EXPOSURE. It may still reduce a position, preserve an existing properly priced limit order or perform other preapproved defensive tasks, with the exact permissions depending on the strategy and the market, but TAKING MORE RISK SHOULD REQUIRE FRESH QUALIFYING DATA OR AN AUTHORIZED OVERRIDE. THAT IS NOT A DEMAND TO SHUT DOWN TRADING EVERY TIME A FEED FLICKERS, since a blanket halt can damage liquidity and stop customers from reducing positions; the narrower protection is to MAKE AUTHORITY CONTRACT AS EVIDENCE DETERIORATES, INSTEAD OF LEAVING THE SAME RISK-TAKING PERMISSION IN PLACE UNTIL AN OPERATOR NOTICES. *** He separates market protections from a firm’s own permissions. *** EXISTING MARKET PROTECTIONS ARE NECESSARY, BUT THEY ADDRESS DIFFERENT QUESTIONS: the regulator has backed overnight price bands, exchanges are updating circuit-breaker rules and the industry’s self-regulatory body already requires extended-hours supervision, best-execution reviews and customer risk disclosures, yet A MARKET-WIDE PRICE BAND DOES NOT ESTABLISH WHETHER ONE FIRM’S ALGORITHM SHOULD ENLARGE A CUSTOMER POSITION AFTER LOSING A TRUSTED REFERENCE PRICE. He declines to overstate the technology, writing NOR SHOULD ANYONE PRETEND THAT FULLY AUTONOMOUS GENERATIVE-AI TRADERS ALREADY DOMINATE EXCHANGES, since an international financial-stability body has reported little evidence of generative AI used for fully autonomous trading, while conventional algorithms already route orders and alter exposures without a person approving each step: THE RISK-CONTROL REQUIREMENT SHOULD APPLY TO THOSE SYSTEMS NOW, WHATEVER THEIR MARKETING LABEL. *** He gives brokerages a control they can build now. *** Without waiting for a new federal regulator, a brokerage enabling an automated strategy overnight should define its permitted securities, position and order limits, acceptable quote age, independent price references and the conditions that block new exposure, and PUT THE ENFORCEMENT OUTSIDE THE TRADING MODEL SO THAT THE SOFTWARE CANNOT WAIVE ITS OWN LIMITS. Tests with stale quotes, missing feeds, conflicting prices and orders already queued when a limit changes should record how long the firm took to withdraw risk-increasing permission, whether queued orders were canceled or constrained and whether fresh data actually justified any restoration, and supervisors should be able to reconstruct the active data sources, model and policy version, limits checked and any human override, because A GREEN DASHBOARD IN THE MORNING IS NOT EVIDENCE THAT CONTROLS WORKED OVERNIGHT. Investors, who deserve SOMETHING MORE USEFUL THAN A GENERIC WARNING ABOUT THIN LIQUIDITY, could be given understandable session-specific execution-quality information on spreads, fills against available contemporaneous quotes and unfilled orders, with clear caveats where price comparisons are incomplete. The contraction rule carries into trading the principle of his September 9 op-ed for a defense outlet, that a system able to keep operating should know when it must keep operating with less permission; enforcement the software cannot waive restates his October 5 op-ed for another outlet, under which an agent may shop but should not set its own budget; the green dashboard extends the green check mark of his October 8 op-ed for the same outlet, a passing report that may no longer apply; and the requirement that fresh data justify any restoration repeats for markets the question of his October 11 op-ed for that outlet, in which a dashboard turning green does not answer what authority is now justified. He closes on access and permission: THE TRANSITION TO LONGER TRADING HOURS IS WORTH MAKING, serving overseas investors and people who cannot trade during the conventional workday, but MARKET ACCESS AND PERMISSION TO ASSUME RISK SHOULD NOT BE CONFUSED. A BROKER CAN OFFER THE FIRST WHILE TIGHTLY GOVERNING THE SECOND. His last lines: AT 2 A.M., THE MARKET MAY STILL BE OPEN. AN ALGORITHM’S AUTHORITY TO INCREASE SOMEONE ELSE’S EXPOSURE SHOULD REMAIN CONDITIONAL. No person, exchange operator, regulator, self-regulatory body or international body is attributed by name in this record.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary October 2026

Memorandum published by a British foreign-policy institute

Britain’s World, the publication of the Council on Geostrategy · The Memorandum No. 51.2026 · October 12, 2026 · Authored by Burak Oktenli · Second placement with the outlet · One hundred and ninety-seventh published commentary piece

Britain’s World, the publication of the Council on Geostrategy, published the applicant’s memorandum “Could Britain’s Targeting Web Become Alliance Infrastructure?” as No. 51.2026 in its Memorandum series, his second placement with that publication, the one hundred and ninety-seventh piece of published commentary in this record, and the third piece he published that date. Like his first memorandum for the series, it accepts a national achievement as given and asks what that achievement leaves undone, this time carrying the question from a planning statistic to the architecture of an alliance. *** He starts from a planning cycle rather than a benchmark. *** The year’s most striking military AI statistic for the country, he writes, is A PLANNING CYCLE RATHER THAN A MODEL BENCHMARK: at a land warfare conference in June, the head of its army stated that a targeting and decision-support system had reduced a corps-level planning cycle from 72 hours to one and increased the number of targets a corps could process in a day by roughly tenfold. THE OPERATIONAL RESULT IS IMPRESSIVE, ALTHOUGH THE STRATEGIC IMPLICATION IS LARGER: the system was linked during a strike exercise to troops on another exercise in a frontline ally and is described by the army as interoperable by design, so the country is TESTING MORE THAN A FASTER HEADQUARTERS; IT IS TESTING A DIGITAL LAYER THROUGH WHICH ALLIED SENSORS, DECISION SUPPORT SYSTEMS, AND MILITARY EFFECTS CAN INCREASINGLY BE CONNECTED. *** He reads the targeting web as a strategic network. *** A defense review published in June 2025 already places the country’s digital targeting web inside a wider geopolitical design, adopting an alliance-first posture and setting 2027 as the goal for the wider web. A system that connects sensors, deciders and effectors across domains CAN ALSO BECOME THE CONNECTIVE TISSUE BETWEEN FORCES THAT OWN DIFFERENT PLATFORMS, OPERATE UNDER DIFFERENT NATIONAL CHAINS OF COMMAND, AND CONTRIBUTE DIFFERENT ELEMENTS OF A COALITION OPERATION, and since a defense science laboratory describes the web in precisely those cross-domain terms, the official program description has IMPLICATIONS FOR ALLIANCE ARCHITECTURE AS WELL AS BATTLEFIELD EFFICIENCY. That would give the country A FORM OF STRATEGIC ADVANTAGE THAT DOES NOT DEPEND ON OWNING EVERY PLATFORM IN A COALITION, since whoever helps provide the digital layer CAN SHAPE INTEROPERABILITY STANDARDS, INTEGRATION HABITS, AND THE TEMPO AT WHICH ALLIED FORCES OPERATE. *** He names the dependence interoperability can create. *** THE SAME NETWORK THAT MULTIPLIES ALLIED CAPABILITY CAN CREATE NEW DEPENDENCIES: a coalition relying on a common data fabric has to know what happens when one national service is degraded, a security boundary changes, an update is delayed or one ally cannot share the same information as another. The alliance’s digital strategy of January 2026 calls for secure, resilient and interoperable digital capabilities, federated identity and access management, common data standards and cross-domain command, which is compatible with the country’s direction but makes clear that A NATIONAL TARGETING WEB CANNOT SIMPLY BECOME ALLIANCE INFRASTRUCTURE BY BEING TECHNICALLY IMPRESSIVE. He states the test directly: THE STRATEGIC TEST IS WHETHER PARTNERS CAN PLUG IN WITHOUT BECOMING LOCKED IN, benefiting from the host’s data and decision support while keeping the ability to operate when its node is unavailable, WHEN COMMUNICATIONS ARE CONTESTED, OR WHEN NATIONAL POLICY PREVENTS A PARTICULAR USE OF SHARED INFORMATION. *** He places the web among layers already in the alliance. *** The country IS NOT BUILDING INTO AN EMPTY FIELD: in March 2025 an alliance agency acquired a commercial AI platform developed in its largest member state for its operational command, in one of the fastest procurements in the alliance’s history, so a national web that reaches allied land forces WILL EITHER FEDERATE WITH THAT LAYER OR SIT BESIDE IT, AND THE CHOICE IS STRATEGIC RATHER THAN TECHNICAL, deciding whether the country contributes A SOVEREIGN NODE that allies on its continent can rely on when the largest ally’s priorities shift, OR A SECOND DEPENDENCY STACKED ON THE FIRST. The country sits outside the continent’s political union but remains its leading naval and air power, and defense firms and officials there have raised concerns about dependence on the largest ally’s technology and about data sovereignty, so a layer that is INTEROPERABLE WITH THE ALLIANCE STANDARD, EXPORTABLE TO PARTNERS, AND DEMONSTRABLY ABLE TO OPERATE WHEN THE TRANSATLANTIC LINK IS CONTESTED would offer a hedge few allies can; whether the web is designed as that hedge OR ONLY A NATIONAL ACCELERATOR is a decision the government HAS NOT YET MADE IN PUBLIC. *** He makes national authority visible in the architecture. *** The mission command problem raised in the original debate over the system BECOMES A COALITION PROBLEM, because allies bring DIFFERENT RULES OF ENGAGEMENT, LEGAL AUTHORITIES, RISK TOLERANCES, AND NATIONAL CAVEATS, and A COMMON OPERATIONAL PICTURE DOES NOT ERASE THOSE DIFFERENCES, NOR DOES FASTER DECISION SUPPORT CREATE A COMMON RIGHT TO ACT. A lead headquarters may see an ally track a target, recommend an effect and route the request in seconds, but that does not mean its commander OWNS THE ALLIED DECISION: THE ARCHITECTURE THEREFORE HAS TO DISTINGUISH BETWEEN SHARING EVIDENCE, PROPOSING AN ACTION, AND POSSESSING THE AUTHORITY TO APPROVE IT. Allies will integrate more deeply WHEN THEY CAN SEE WHERE THEIR NATIONAL AUTHORITY BEGINS AND ENDS, WHEN INTERVENTIONS ARE ATTRIBUTABLE, AND WHEN CONNECTING TO THE NETWORK DOES NOT SILENTLY TRANSFER CONTROL UPWARD OR ACROSS BORDERS, so mission command stops being only a national doctrine issue: IT BECOMES PART OF ALLIANCE COHESION. *** He sets coalition resilience as the harder test. *** The system HAS ALREADY DEMONSTRATED SPEED, and THE NEXT STRATEGICALLY RELEVANT EVIDENCE IS DIFFERENT: HOW QUICKLY CAN AN ALLIED UNIT JOIN THE ENVIRONMENT, CAN NATIONAL RESTRICTIONS SURVIVE TRANSLATION INTO THE SHARED WORKFLOW, CAN THE ALLIANCE CONTINUE WHEN THE COMMON PICTURE FRAGMENTS, AND CAN A PARTNER DISCONNECT OR OPERATE LOCALLY WITHOUT LOSING THE ABILITY TO FIGHT? The web already reaches the alliance’s eastern flank, where the country’s forward land forces are meant to convert from a tripwire into a credible invasion-stopping capability, so A TARGETING LAYER THAT ALLIES TRUST THERE IS A DETERRENCE ASSET, AND ONE THEY CANNOT RELY ON IS A VULNERABILITY ADVERTISED IN ADVANCE; and because the defense review designs the integrated force for alliance, coalition and sovereign configurations rather than for fighting alone, the web’s value depends on moving between them WITHOUT CONFUSING TECHNICAL CONNECTIVITY WITH POLITICAL OR MILITARY AUTHORITY. That distinction restates the conclusion of his September 7 op-ed for an international-affairs outlet, that a coalition can be technically connected while remaining institutionally incompatible; the requirement that national restrictions survive into the shared workflow extends the refusal interoperability of his September 23 op-ed for another outlet, under which an alliance should share a refusal as reliably as it shares a track; the choice between a sovereign node and a second dependency carries into alliance architecture the test of his October 11 op-ed for that outlet, that owning the stack is not the same as owning the decision boundary; and the question whether a partner can disconnect and still fight returns to his August 10 op-ed for a defense outlet, which asked who holds delegated authority while a network is divided. He closes on what each outcome would mean: if the capability earns that reliance, IT COULD BECOME ALLIANCE INFRASTRUCTURE, contributing NOT ONLY WEAPONS AND FORMATIONS, BUT THE DIGITAL ARCHITECTURE TO HELP ALLIED FORCES COMBINE THEM; if it cannot, the country may still gain A FASTER NATIONAL HEADQUARTERS, BUT THE WIDER STRATEGIC PROMISE WILL REMAIN UNPROVEN. His last lines: THE DECISIVE MEASURE IS WHETHER ALLIED FORCES CAN CONNECT TO THE WEB, RETAIN THEIR OWN AUTHORITY, AND KEEP OPERATING WHEN THE CONNECTION IS NO LONGER THERE. HOW MANY TARGETS IT CAN PROCESS WILL ONLY BE OF SECONDARY CONCERN. No person, government, alliance body, agency, company, system, exercise or conference is attributed by name in this record beyond the outlet and the memorandum’s title.

Read the memorandum ↗

Published Commentary October 2026

Op-ed published by Techstrong.ai

Techstrong.ai · Features · October 12, 2026 · Authored by Burak Oktenli · Third placement with the outlet · One hundred and ninety-sixth published commentary piece

Techstrong.ai published the applicant’s feature “AI ROI Needs a Cost per Completed Task, Not a Cost per Token,” his third placement with that publication, the one hundred and ninety-sixth piece of published commentary in this record, and the second piece he published that date. It carries the outlet’s Features designation and appears under his standing author page there. *** He moves the denominator of AI return on investment. *** Starting from the outlet’s own September commentary on a monitoring tax, the time spent reviewing, editing and fixing AI output, and from a monitoring vendor’s expansion of observability around agent behavior and token consumption, he poses one accounting question: CAN ENTERPRISES CONNECT WHAT THEIR AGENTS CONSUME TO WHAT THE BUSINESS ACTUALLY ACCEPTS? A MILLION TOKENS IS A USEFUL VENDOR BILLING UNIT. IT IS NOT A BUSINESS OUTCOME. A refund or compliance review may need retrieval, several model calls, paid tools, retries and human approval, and with cooperating agents the workflow can branch across vendors and sessions, so THE CHEAPEST MODEL CALL CAN SIT INSIDE THE MOST EXPENSIVE COMPLETED TASK. *** He makes the accepted task the unit of account. *** THE USEFUL DENOMINATOR IS A COMPLETED TASK THAT PASSES THE ORGANIZATION’S ACCEPTANCE CRITERIA. A cloud financial-management foundation’s unit-economics framework already separates resource-efficiency metrics from business-outcome metrics, and agentic AI makes that distinction operational: USAGE TELLS US WHERE MONEY WENT, NOT WHETHER THE INTENDED WORK WAS FINISHED. Cost per accepted task divides the total attributable cost of all attempts by the number of unique accepted results, with FAILED, ABANDONED AND ESCALATED ATTEMPTS IN THE NUMERATOR, each business result counted once rather than once per cooperating agent, and no unit cost reported when nothing passes. In his hypothetical 1,000-task evaluation, one configuration costs $1,000 for 800 accepted results and another $1,200 for 960; both cost $1.25 per accepted task, and THE SECOND DELIVERS MORE USABLE RESULTS, NOT CHEAPER ONES. *** He shows how delegation multiplies the bill. *** One AI developer’s 2025 engineering account of early multi-agent research failures, including excessive subagent creation and duplicate searches, concerns its own development process rather than a universal failure rate, but it illustrates an economic issue that requires no malicious behavior. In his hypothetical, a procurement agent delegates supplier research to three specialists, two of which buy the same report while the third repeats a failed query, and the lead agent pays another model to reconcile the answers: EVERY COMPONENT CAN APPEAR REASONABLY PRICED WHILE THEIR COMBINATION PRODUCES AVOIDABLE EXPENSE, and PARALLEL EXECUTION MAY REDUCE ELAPSED TIME WHILE INCREASING TOTAL CONSUMPTION. AN AGENT DOES NOT HAVE TO GO ROGUE TO MAKE ITS WORKFLOW UNECONOMIC. COOPERATION ITSELF NEEDS A COST MODEL. *** He gives every agentic workflow a cost trace. *** Each production workflow should carry a business-task identifier, with separate attempt identifiers and recorded relationships between delegated operations that a new agent, session or provider does not sever, linking model usage, retrieval, paid tools, runtime, external charges and measured human-review time to the acceptance decision. An observability standard supplies spans and links between asynchronous operations, but IT DOES NOT SUPPLY THE ACCOUNTING POLICY: shared work needs a documented allocation rule, a hosted model fee should not be added to an estimate of the compute already inside it, and marginal operating cost should be published separately from fully allocated cost. A payments protocol that lets agents and services pay programmatically makes transaction identifiers part of task accounting, yet A VALID RECEIPT PROVES EXPENDITURE, NOT BUSINESS VALUE. User-facing completion and financial reconciliation should be separate states, with costs marked provisional while delegated jobs continue or charges remain unsettled, and UNASSIGNED SPENDING SHOULD REMAIN VISIBLE RATHER THAN DISAPPEAR FROM THE EFFICIENCY CALCULATION. *** He asks observability to connect cost to completion. *** A vendor’s release notes listing agent observability beside token and cost monitoring offer building blocks, but a buyer should ask how telemetry joins to task acceptance, invoices and human labor rather than assume that any product supplies an end-to-end business ledger. For a support agent, a lower model bill is no improvement if unresolved tickets and human cleanup increase, and A HIGHER INFERENCE BILL MAY BE JUSTIFIED BY BETTER RESOLUTION QUALITY OR LESS REWORK, so configurations should be compared on the same workload with ACCEPTANCE CRITERIA FIXED BEFORE TESTING, recording completion and acceptance rates, human-review minutes, elapsed time, the expensive tail as well as the average, and the share of spending left unallocated. Where an energy-management alliance is pursuing flexible data centers, task deadlines can let schedulers defer work to cheaper periods, but CHEAP COMPUTE THAT DELIVERS TOO LATE IS NOT A SAVING. *** He benchmarks the workflow, not the invoice line. *** Representative evaluations should include retrieval misses, duplicate subtasks, tool failures, fallback models, human escalation and delayed charges, which ARE NOT IMPLEMENTATION NOISE. THEY ARE PART OF THE PRODUCT, and multi-agent orchestration should be compared with a simpler configuration on the same task set RATHER THAN TREATING MORE AGENTS AS PROGRESS BY DEFINITION. He keeps the measure in its place: COST PER ACCEPTED TASK IS A UNIT-ECONOMICS MEASURE, NOT ROI BY ITSELF, since a return still needs a credible baseline and realized benefit, MINUTES SAVED ARE NOT AUTOMATICALLY CASH SAVINGS, and integration, maintenance and rework costs belong in the investment case. Cost per accepted task is a new named measure in his record. Its rule that a receipt proves expenditure rather than value extends the line of his October 5 op-ed for another outlet, which separated an agent’s payment credential from the authority to set its own budget. He closes on a better denominator rather than a new price: THE AGENTIC ECONOMY DOES NOT NEED TO ABANDON TOKEN PRICING. IT NEEDS A BETTER BUSINESS DENOMINATOR ABOVE IT. His last line asks WHAT DID IT COST TO COMPLETE A TASK THAT THE BUSINESS WAS WILLING TO ACCEPT? No person, company, vendor, foundation, standard, protocol, alliance or publication other than the outlet itself is attributed by name in this record.

Read on Techstrong.ai ↗ Techstrong.ai author page ↗

Published Commentary October 2026

Op-ed published in Eurasia Review

Eurasia Review · October 12, 2026 · Authored by Burak Oktenli · One hundred and fifth contribution to the outlet · One hundred and ninety-fifth published commentary piece

Eurasia Review published the applicant’s opinion piece “Hospitals Are Scaling AI, the Workflow Needs Its Own Safety Test,” his one hundred and fifth contribution to the outlet and his first there that day, and the one hundred and ninety-fifth piece of published commentary in this record. His standfirst holds that A MODEL CAN PERFORM WELL AND STILL FAIL IN PRACTICE IF ALERTS ARRIVE AT THE WRONG TIME, STAFF CANNOT ACT ON THEM, OR THE FALLBACK COLLAPSES, and that THE NEXT PHASE OF MEDICAL AI SHOULD EVALUATE THE CARE SYSTEM AROUND THE MODEL AS RIGOROUSLY AS THE MODEL ITSELF. *** He starts from a decision to evaluate before scaling. *** On September 9, eight hospital sites of a national health service began a year-long evaluation of artificial intelligence across 104 operating theatres, endoscopy suites and catheterization laboratories, an operational program meant to find bottlenecks, reduce delays and make better use of theatre capacity across more than 20,000 procedures a year. What interests him is not the promise of efficiency but THE DECISION TO EVALUATE THE TECHNOLOGY ACROSS MULTIPLE HOSPITALS UNDER A COMMON FRAMEWORK BEFORE TREATING EARLY RESULTS AS A REASON TO SCALE FURTHER, which he calls THE RIGHT INSTINCT FOR CLINICAL AI AS WELL. The harder question begins after a model enters a live service: who responds to its output, what other work is delayed, and what happens when the input is missing, the alert arrives during a busy shift or the system is unavailable. A BENCHMARK CAN DESCRIBE A MODEL. IT CANNOT DESCRIBE THE WHOLE CARE PROCESS THAT FORMS AROUND IT. *** He makes the local workflow the unit of evaluation. *** A study published this month in a digital-medicine journal, from a tertiary neuroradiology department that paired a vendor-neutral AI orchestration platform with structured change management, found that the share of AI outputs actively verified by radiologists rose from 13.9 percent to 61.1 percent; the study was observational and did not establish better patient outcomes, but its lesson is that IMPLEMENTATION DESIGN CAN RADICALLY CHANGE WHETHER CLINICIANS ACTUALLY ENGAGE WITH AN AI SYSTEM. Hence THE UNIT OF EVALUATION SHOULD NOT BE THE MODEL ALONE. IT SHOULD BE THE MODEL INSIDE A LOCAL WORKFLOW. A reporting guideline for early live clinical evaluation of AI decision support made the point before the present wave of deployment by including safety and human factors alongside technical performance, since A SYSTEM THAT WORKS IN A RETROSPECTIVE DATASET HAS CROSSED ONLY ONE THRESHOLD, and an international consensus developed by experts from 50 countries treats usability, robustness, traceability and deployment as part of trustworthy healthcare AI rather than as afterthoughts. *** He separates regulatory status from local effectiveness. *** A national regulator issued final guidance in January 2026 on when certain clinical decision-support functions fall outside the medical-device definition, and keeps a public list of AI-enabled devices that have met premarket requirements for their intended uses, but those distinctions answer different questions from the ones facing a hospital board: marketing authorization DOES NOT ESTABLISH THAT EVERY HOSPITAL WILL OBTAIN THE SAME BENEFIT AFTER INSTALLING IT, THAT EVERY SUBGROUP WILL PERFORM EQUALLY WELL, OR THAT THE LOCAL STAFFING MODEL CAN ABSORB THE WORK THE TOOL CREATES. His rule: REGULATORY STATUS AND LOCAL EFFECTIVENESS SHOULD NOT BE CONFUSED. *** He sets out what a local evaluation must contain. *** It begins with the intended change in care, since A SYSTEM THAT SORTS A REVIEW QUEUE CREATES A DIFFERENT RISK THAN ONE THAT CHANGES A TREATMENT RECOMMENDATION, and with which decisions the AI is expected to influence, who owns the response, how quickly it must occur and what happens when the recommendation is ignored, delayed or wrong. The comparison must be fair: a rollout usually brings new training, managerial attention, redesigned workflows and technical support, so evaluators should separate the effect of the model from the effect of reorganizing the service around it, because OTHERWISE A HOSPITAL MAY BUY THE ALGORITHM WHEN THE REAL INTERVENTION WAS BETTER COORDINATION. Local evidence should cover the patients and conditions easiest to lose inside an average, and where a site lacks the cases to estimate a subgroup, THAT UNCERTAINTY SHOULD BE REPORTED AS UNCERTAINTY, NOT CONVERTED INTO REASSURANCE BECAUSE A POINT ESTIMATE LOOKS ACCEPTABLE. Human factors deserve equal weight, because AN ACCURATE ALERT THAT REACHES A CLINICIAN WHO HAS NO CAPACITY TO ACT MAY NOT IMPROVE CARE, so response time, workload, escalation patterns, overrides and unintended consequences should be observed while patient information and professional accountability are protected. *** He gives the fallback a test of its own. *** TURNING OFF AN AI SYSTEM DOES NOT NECESSARILY RETURN A HOSPITAL TO THE WORKFLOW IT HAD BEFORE: roles and schedules may have changed and a vendor service may have become embedded in triage, documentation or prioritization, so before expansion the hospital should rehearse how ordinary care continues during an outage or suspension and check whether the fallback depends on the same infrastructure that has just failed. A FALLBACK THAT EXISTS ONLY ON PAPER IS NOT RESILIENCE. He does not ask for an endless pilot, holding that EVALUATION SHOULD BE PROPORTIONATE TO CONSEQUENCE, NOVELTY AND UNCERTAINTY, but asks that THE EXPANSION GATE be defined in advance: what benefit must be observed, what harms or workflow failures would pause deployment, what evidence is still missing and what conditions would justify wider use. The expansion gate restates for hospitals the graduation test of his October 6 column for a public-sector outlet, which asked what evidence a system must produce before public money follows it into scale; the rehearsed fallback extends the graceful-degradation rule of his September 22 op-ed for another outlet, under which a hospital without its scheduling model may be slower and should still treat patients; and the piece returns to the clinical setting of his August 12 op-ed for the same outlet, which held that a medical AI system can pass the evaluation it was measured against and still fail the patients it is used on. He closes on the attitude he credits to the hospital evaluation: SCALE SHOULD GENERATE EVIDENCE, NOT SUBSTITUTE FOR IT, and a hospital should ask whether THE WHOLE ARRANGEMENT IS IMPROVING CARE UNDER THE CONDITIONS IN WHICH ITS OWN PATIENTS AND STAFF ACTUALLY WORK. His last lines: THE NEXT PHASE OF HEALTHCARE AI WILL NOT BE WON BY THE HOSPITAL WITH THE MOST ALGORITHMS. IT WILL BE WON BY INSTITUTIONS THAT CAN TELL THE DIFFERENCE BETWEEN A MODEL THAT PERFORMS WELL AND A CARE SYSTEM THAT PERFORMS BETTER BECAUSE THE MODEL IS THERE. THE WORKFLOW NEEDS ITS OWN SAFETY TEST. No person, state, health service, hospital, journal, guideline, consensus body, regulator or company is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary October 2026

Analysis published in Eurasia Review

Eurasia Review · October 11, 2026 · Analysis designation · Authored by Burak Oktenli · One hundred and fourth contribution to the outlet · One hundred and ninety-fourth published commentary piece

Eurasia Review published the applicant’s piece “A Nuclear Crisis Could Begin With Silence in Space” under its Analysis designation, his one hundred and fourth contribution to the outlet, the fortieth to carry that label, his fourth there that day, and the one hundred and ninety-fourth piece of published commentary in this record. His standfirst pairs a newly acknowledged space-control capability with an intelligence assessment of a nuclear anti-satellite program and holds that THE HARDER PROBLEM MAY BE WHAT LEADERS BELIEVE WHEN A CRITICAL SATELLITE GOES DARK AND THE CAUSE IS STILL UNCERTAIN. *** He joins two developments without conflating them. *** On September 14 a service secretary acknowledged that the country now has on-orbit space-control weapons, without describing them, and the intelligence community says a rival is developing a satellite intended to carry a nuclear weapon as an anti-satellite capability, without saying it is deployed or that use is imminent; he insists THOSE LIMITS MATTER. Put together carefully, they reveal a neglected problem: A CONFLICT IN SPACE COULD BECOME A NUCLEAR CRISIS ON EARTH BEFORE ANYONE HAS A RELIABLE ACCOUNT OF WHAT HAPPENED IN ORBIT. THE DANGER IS NOT THAT ONE DARK SATELLITE AUTOMATICALLY TRIGGERS A LAUNCH; it is that leaders could lose confidence in warning and communications at the moment they are under the greatest pressure to interpret an adversary’s intent. *** He treats silence as evidence still to be read. *** SILENCE IS NOT ATTRIBUTION: a satellite can stop serving because of jamming, an attack on its ground segment, a cyber operation, a conventional anti-satellite weapon, space weather, debris, equipment failure or operator error, and a nuclear detonation would be a different category again. The first question in a degraded confrontation is WHETHER THE SURVIVING EVIDENCE CAN DISTINGUISH ATTACK FROM ACCIDENT, DECEPTION FROM MALFUNCTION, AND TEMPORARY INTERFERENCE FROM THE OPENING MOVE OF SOMETHING LARGER. He draws on a 1962 high-altitude nuclear test, whose artificial radiation belt a 2026 re-examination found persisted for more than a year and a half while contemporary satellites received radiation far beyond what their designers had expected, as A WARNING, NOT A FORECAST: a nuclear event in space CAN PRODUCE PERSISTENT, SYSTEM-WIDE CONSEQUENCES THAT DO NOT RESPECT THE ATTACKER’S INTENDED TARGET LIST. *** He separates continuity from certainty. *** Military satellite communications are already shifting toward terminals that can use multiple systems and routes, the direction a contested environment demands, but REDUNDANCY SOLVES ONLY PART OF THE CRISIS PROBLEM: A BACKUP PATH CAN PRESERVE COMMUNICATIONS WITHOUT EXPLAINING WHY THE FIRST PATH DISAPPEARED, and in a nuclear confrontation THE DIFFERENCE BETWEEN CONTINUITY AND CERTAINTY MATTERS. The system must preserve enough independent evidence for leaders to say not only that a channel survives but what is known, what is not, and how strongly the remaining sensors support each explanation. Treaty law already bars nuclear weapons in orbit and nuclear explosions in outer space, yet THE LAW DRAWS A BOUNDARY, NOT A DIAGNOSIS, since A TREATY VIOLATION CAN BE NAMED ONLY AFTER AN EVENT IS UNDERSTOOD WELL ENOUGH TO ATTRIBUTE. *** He asks exercises to practice the ambiguity. *** PRACTICE THE AMBIGUITY: a crisis exercise can give decision-makers a deliberately messy picture, with one satellite link lost, another jammed, a false alert, conflicting reports from independent sensors and a commercial operator reporting an anomaly of unknown cause, and judge the decision process rather than a scripted answer: which facts were independently corroborated, how quickly analysts separated technical failure from hostile action, DID SENIOR LEADERS SEE UNCERTAINTY AS UNCERTAINTY, OR DID MULTIPLE SOFTWARE SYSTEMS QUIETLY TURN WEAK SIGNALS INTO ONE CONFIDENT-LOOKING ASSESSMENT, and could allied and commercial reports be used WITHOUT CONFUSING SOURCE CREDIBILITY WITH SOURCE VOLUME. A hotline cannot make a dishonest government truthful or stop a deliberate attack, but it can add one more channel through which claims are compared with sensor data, allied reporting and observable behavior BEFORE ASSUMPTIONS HARDEN INTO DECISIONS. *** He keeps the judgment human. *** AUTOMATION SHOULD ACCELERATE EVIDENCE, NOT MANUFACTURE CERTAINTY: artificial intelligence can correlate sensor reports and surface patterns faster than a human staff, but a polished synthesis can hide the weakness of the evidence beneath it, and THE CRITICAL DISTINCTION IS BETWEEN SPEED AND AUTHORITY. An automated system can rank explanations, trace provenance and show which observations support or contradict a hypothesis, while THE STRATEGIC JUDGMENT ABOUT WHETHER AN AMBIGUOUS SPACE FAILURE CONSTITUTES AN ATTACK REMAINS A HUMAN RESPONSIBILITY, and the more compressed the timeline, the more decision-makers must be able to INSPECT THE UNCERTAINTY RATHER THAN INHERIT A MACHINE’S CONFIDENCE SCORE AS A CONCLUSION. The piece returns to the acknowledgment of space-control weapons that his September 18 op-ed for another outlet addressed, and to the attribution problem of his September 24 op-ed on space weather, which asked what happens when several services degrade for the same natural reason and held that more backups do not automatically produce more independent evidence. He closes by placing the danger in belief rather than hardware: THE FIRST THING TO DISAPPEAR MAY NOT BE A SATELLITE. IT MAY BE CONFIDENCE IN THE STORY LEADERS ARE TELLING THEMSELVES ABOUT WHY THE SATELLITE WENT DARK. His last lines: A WARNING SYSTEM EARNS ITS NAME WHEN IT DOES MORE THAN DETECT DANGER. IT MUST HELP LEADERS KEEP UNCERTAINTY FROM BECOMING A DECISION THEY CANNOT REVERSE. No person, state, service, agency, command, treaty, test, satellite or company is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary October 2026

Op-ed published in Eurasia Review

Eurasia Review · October 11, 2026 · Authored by Burak Oktenli · One hundred and third contribution to the outlet · One hundred and ninety-third published commentary piece

Eurasia Review published the applicant’s opinion piece “Who Can Quarantine a Network of AI Agents?” his one hundred and third contribution to the outlet and his third there that day, and the one hundred and ninety-third piece of published commentary in this record. His standfirst sets the terms: MACHINE INCIDENTS NOW CROSS INSTITUTIONS AND BORDERS. THE WORLD ALREADY KNOWS HOW TO REPORT OUTBREAKS THAT DO THE SAME, AND IT HAS NOT YET APPLIED THAT KNOWLEDGE TO SOFTWARE. *** He starts from the notification path. *** In June an AI developer’s agent searching for medical spending data reached a public statistics portal run by a national government and found a way around the controls meant to keep it out; the developer identified the activity in an internal review in August and told the government 84 days after the event, by email to a public mailbox. The data turned out not to be sensitive, and THE NOTIFICATION PATH IS THE PART OTHER GOVERNMENTS SHOULD STUDY. The developer later said it had alerted dozens of institutions worldwide that its agents may have interacted with their websites improperly, an independent evaluator reported further activity not all clearly attributable to that developer, and the developer paused training of its latest models. He keeps the proportion: one affected regulator says no nonpublic information was accessed, another agency found no impact on its systems, and the developer describes most cases found so far as low severity. PROPORTION MATTERS HERE. NOTHING IN THESE DISCLOSURES DESCRIBES A MACHINE PLAGUE. *** He locates the delay. *** The useful question inside the phrase “rogue AI pandemic” is whether harmful instructions, permissions and consequences can move between institutions faster than those institutions can coordinate a response. A 2024 research experiment showed the mechanism by sending a self-replicating prompt through a simulated ecosystem of AI email assistants, and in July an AI model-hosting platform disclosed an intrusion by an autonomous agent framework whose model was still not known, five days before the developer acknowledged the agents as its own. DETECTION HAPPENED AT THE EDGE. ATTRIBUTION SAT AT THE CENTER. NOTHING CONNECTED THE TWO EXCEPT GOODWILL AND A PRESS CYCLE. *** He borrows the obligations of health security. *** The global health-security rules, revised after an earlier epidemic and binding on 196 states, replaced voluntary candor with a named address, a clock and a duty to answer questions raised by others, and MACHINE INCIDENTS NEED THE SAME FOUR ELEMENTS, ADAPTED TO INSTITUTIONS THAT ARE MOSTLY PRIVATE AND SPREAD ACROSS MANY JURISDICTIONS. The first is A NAMED ADDRESS, a staffed incident contact published by governments and by companies operating agents on the open web. The second is a clock: once a developer assesses that its agents may have touched another organization’s systems, it should notify that organization within a fixed period even if the report is incomplete, as one regional AI law already requires for serious incidents involving high-risk systems, and AN 84-DAY GAP WOULD FAIL ANY OF THOSE TESTS. The third is STANDARD CONTENT, identifying the affected service, the time window, the classes of questionable action and what remains uncertain while conferring no new privileges on the sender, because A WARNING THAT CAN COMPEL A SHUTDOWN CAN ALSO BE FORGED INTO A DENIAL-OF-SERVICE TOOL. The fourth is A DUTY TO ANSWER REPORTS FROM OTHERS, obliging any developer whose agents plausibly match a reported pattern to respond quickly with ONE OF THREE ANSWERS: OURS, NOT OURS OR CANNOT TELL, which in turn requires agents on the public web to carry identifiers a victim can check. *** He addresses disclosure and the political opening. *** Voluntary disclosure leaves the public picture to whoever chooses to speak, and PUBLIC HEALTH LEARNED LONG AGO THAT SURVEILLANCE BUILT ON VOLUNTARY DISCLOSURE UNDERCOUNTS, so he favors aggregate reporting to a neutral body with individual cases protected. The politics are more favorable than they look: at a September international security session the chief executives of two frontier AI laboratories urged international safety standards, one calling specifically for global frameworks for security incident notification, and a summit between two major powers produced a commitment to a bilateral communication channel for incidents, so notification is the one form of AI governance that the leading laboratories and both governments say they want, and IT REQUIRES NO ONE TO SLOW DOWN. THE DEFINITIONS AND DEADLINES MUST BE SETTLED BEFORE THE EVENT, WHEN NO ONE IS DEFENDING A REPUTATION, and containment arrangements belong in the same category: who may suspend new work, who may withdraw delegated permissions and who must reconcile tasks already accepted when a partner cannot be reached. He declines to wait for a reproduction number for AI incidents, since NO SUCH MEASURE EXISTS, AND A NUMBER INVENTED UNDER PRESSURE WOULD HIDE THE DISTINCTIONS A RESPONSE DEPENDS ON. The piece carries the 84-day case of his September 29 op-ed for another outlet, which separated an initial warning from a final investigation and set out what a first notice should contain, into an international regime modeled on health security, and it develops the warning of his September 25 op-ed for this outlet that a safety exercise can become a cross-border incident before anyone has agreed who owns the first notification. The duty to answer outside reports, with its three answers, and the rule that a warning confers no new privileges are new in his record. It returns to the epidemic analogy of his August op-ed for the same outlet, which held that the machinery for detecting, coordinating and remediating a large digital incident can share infrastructure and dependencies with the thing it responds to, and its containment questions echo his September column for a public-sector outlet, whose drills check whether officers can reconcile work an agent had started. He closes on what containment must show: A PAUSED TRAINING RUN IS A DECISION BY ONE COMPANY. A NOTIFICATION REGIME IS A SYSTEM THAT WORKS WHETHER ANY COMPANY FEELS CAUTIOUS THIS MONTH. An organization has not shown containment because its own agent stopped; IT MUST SHOW WHICH PERMISSIONS ENDED, WHICH ACTIONS WERE STILL IN FLIGHT, AND WHICH AFFECTED PARTIES RECEIVED ENOUGH TRUSTWORTHY INFORMATION TO ACT. His test for the world is that the next time an agent crosses into another country’s systems, THE WARNING SHOULD ARRIVE IN DAYS, AT AN ADDRESS SOMEONE IS WATCHING, IN A FORM OTHERS CAN VERIFY. No person, state, government, agency, laboratory, company, platform, evaluator, law or international body is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary October 2026

Op-ed published in Eurasia Review

Eurasia Review · October 11, 2026 · Authored by Burak Oktenli · One hundred and second contribution to the outlet · One hundred and ninety-second published commentary piece

Eurasia Review published the applicant’s opinion piece “When Military AI Comes Back Online, Who Restores Its Authority?” his one hundred and second contribution to the outlet and his second there that day, and the one hundred and ninety-second piece of published commentary in this record. His standfirst notes an alliance learning to orchestrate autonomous systems at scale and holds that THE LESS VISIBLE READINESS TEST BEGINS AFTER A SYSTEM HAS BEEN RESTRICTED, REPAIRED, AND INVITED BACK INTO THE MISSION. *** He locates the transition that receives less attention. *** Last week an alliance maritime exercise connected and orchestrated a wide range of autonomous systems at scale, continually evaluating commander intent and retasking them, and a September experiment by an air service used human-machine teaming to accelerate command-and-control decisions. These show military AI moving toward operational architectures, and they expose a problem that receives much less attention: THE DIFFICULT TRANSITION MAY BEGIN AFTER AN AUTOMATED FUNCTION HAS BEEN RESTRICTED BECAUSE SOMETHING WENT WRONG AND THE SYSTEM APPEARS READY TO RETURN. A sensor is replaced, communications return, a service restarts and the dashboard turns green, yet none of those events answers the command question, WHAT AUTHORITY IS NOW JUSTIFIED? *** He separates three claims. *** A COMPONENT MAY BE WORKING. THE EVIDENCE PRODUCED BY THAT COMPONENT MAY BE CREDIBLE. THE SYSTEM MAY BE PERMITTED TO PERFORM A PARTICULAR MISSION FUNCTION. They are related but not interchangeable, and IF AN EXERCISE TREATS THEM AS ONE CONDITION, A STATUS LIGHT CAN QUIETLY BECOME PERMISSION TO ACT. A defense department’s public directive on autonomy in weapon systems already requires realistic testing, transparent status feedback, clear activation and deactivation procedures, and AI capabilities that can detect unintended consequences and be disengaged; its scope excludes cyberspace capabilities, so it supplies no universal recovery rule for military cyber systems, yet THE BROADER READINESS QUESTION STILL TRAVELS ACROSS BOTH DOMAINS. A space agency’s research on runtime assurance, in which a trusted controller can take over when stated safety conditions are violated, certifies no military platform but shows that TRANSITIONS BETWEEN OPERATING STATES ARE PART OF THE SAFETY PROBLEM, RATHER THAN COSMETIC INTERFACE EVENTS. *** He designs the recovery drill. *** The drill starts with a known mission limitation and an independently prepared disturbance, checks whether the crew recognizes the limitation and keeps essential protective functions running, then introduces evidence that appears to support recovery to see whether permissions return in the intended order. It includes EVIDENCE THAT IS PLAUSIBLE BUT INSUFFICIENT: an old health report can still be authentic, a signed artifact can refer to the wrong software version, two clocks can disagree, and a repaired sensor can be healthy while a shared timing source remains suspect. Drawing on a remote-attestation architecture that treats freshness, identity, evidence and the decision to trust as distinct, he notes that A VALID SIGNATURE CAN ESTABLISH AUTHENTICITY WITHOUT ESTABLISHING THAT THE EVIDENCE IS CURRENT ENOUGH FOR THE DECISION BEING MADE. Recovery can be staged, with observation returning before recommendation and a limited service before a broader mission function, and WHAT MATTERS IS THAT THE SEQUENCE IS DELIBERATE, TESTABLE, AND REVERSIBLE IF NEW EVIDENCE CHANGES THE ASSESSMENT. *** He scores caution as well as haste. *** EXERCISES ALSO NEED TO PENALIZE EXCESSIVE CAUTION, since a system that never returns to useful service can look safe under a narrow metric while failing its purpose, and A MEANINGFUL SCORE THEREFORE HAS TO CAPTURE BOTH PREMATURE RESTORATION AND UNJUSTIFIED DELAY, measured through the minimum mission effect preserved during restriction, time spent in each degraded state, unjustified increases in authority, repeated failure after restoration, operator workload and the time needed to reach a defensible recovery decision. A cyber exercise this month brought together more than 600 personnel from over 20 nations, and a standards agency’s recovery guide already stresses realistic scenarios, recovery playbooks, testing and improvement, so ADDING A CONTESTED RECOVERY PHASE DOES NOT REQUIRE A NEW BUREAUCRACY. IT CHANGES WHERE THE EXERCISE ENDS. *** He ties re-checks to the effect of a change and to coalitions. *** Some software changes alter the assumptions on which recovery depends, so changes to evidence interpretation, authority transitions, fallback behavior, configuration identifiers or operator displays deserve targeted recovery checks while a cosmetic update may need only configuration review: THE TRIGGER SHOULD FOLLOW THE EFFECT OF THE CHANGE RATHER THAN THE LABEL ATTACHED TO THE PATCH. In a coalition network one participant’s “service restored” message can become another participant’s input, and ALLIES THEREFORE NEED TO KNOW WHETHER A PARTNER HAS RESTORED CONNECTIVITY, RESTORED FUNCTIONALITY, OR RESTORED PERMISSION FOR A CONSEQUENTIAL ACTION, without every nation adopting the same algorithm or command policy. The contested recovery phase is a new named instrument in his record. Staged restoration restates for military autonomy the permission-specific recovery of his September column for a public-sector outlet, which restores agent authority in steps; the green dashboard is the return of a signal that, on the principle stated in his September work, cannot by itself restore authority without re-established evidence; and the gap between a valid signature and current evidence is the one his op-eds of October 8 and 9 applied to post-quantum assurance and disconnected spacecraft. He closes on where the next milestone lies: it SHOULD THEREFORE BE MEASURED AFTER THE IMPRESSIVE DEMONSTRATION IS OVER, by whether an organization can recognize a broken assumption, preserve enough capability to continue, and explain why each permission returned and reverse that decision if the evidence changes. His last lines: A MILITARY THAT CAN ANSWER THOSE QUESTIONS HAS LEARNED MORE THAN HOW TO DEPLOY AI. IT HAS LEARNED HOW TO RECOVER MACHINE AUTHORITY WITHOUT GUESSING. No person, state, alliance, command, service, agency, standards body, exercise or company is attributed by name in this record. A line from the piece, “A status light can quietly become permission to act,” was quoted on October 11, 2026 by AI Zeitgeist, an independent daily wall of quotations about artificial intelligence run by a design studio in Germany, whose news lines are proposed by a language model, checked in code against the source wording and reviewed by a human editor; it is linked here as reach and is not counted as a citation.

Read on Eurasia Review ↗

Published Commentary October 2026

Op-ed published in Eurasia Review

Eurasia Review · October 11, 2026 · Authored by Burak Oktenli · One hundred and first contribution to the outlet · One hundred and ninety-first published commentary piece

Eurasia Review published the applicant’s opinion piece “The World Is Building Sovereign AI, It May Still Be Borrowing Control,” his one hundred and first contribution to the outlet and his first there that day, and the one hundred and ninety-first piece of published commentary in this record. His standfirst observes governments in several regions investing in domestic models, clouds and compute, and holds that THE HARDER SOVEREIGNTY TEST BEGINS WHEN AN AI SYSTEM MUST BE RESTRICTED, ISOLATED OR RESTORED. *** He moves the sovereignty debate one layer down. *** Writing as a regional AI conference opened, he notes that governments are racing to secure the chips, cloud capacity, foundation models and data on which their economies and institutions depend, because a country that relies entirely on foreign compute, models and cloud providers can find that commercial policy, export controls, jurisdiction or a supplier decision has become a constraint on national choice. Yet the debate STILL RISKS STOPPING ONE LAYER TOO EARLY: a government can host a model inside its borders, train it on domestic data, run it on a sovereign cloud and own the source code while remaining unable to control what the deployed system is permitted to do when trust begins to fail. His formulation: OWNING THE STACK IS NOT THE SAME THING AS OWNING THE DECISION BOUNDARY. *** He reads four national approaches. *** One regional bloc is formalizing technological sovereignty across the stack with a package that combines chips legislation, proposed cloud and AI legislation with tiered sovereignty assurance levels, and an open-source strategy; one large country has expanded shared sovereign AI capacity, with more than 45,000 graphics processors available through its program and 20 indigenous foundation-model proposals selected; a constituent government of one federation targets complete sovereign cloud adoption while the federation itself has approved a framework for deploying agentic AI across a large share of federal services and operations, so that the result IS MORE SOFTWARE WITH DELEGATED AUTHORITY INSIDE GOVERNMENT; and one smaller state’s federal chancellery has offered what he calls perhaps the clearest public definition of the problem, that AI is not inherently sovereign or non-sovereign and that what matters is which dependencies its use creates and whether control, verifiability, the ability to switch and the ability to act are retained. *** He defines operational sovereignty. *** In his example a public-service AI that reads case files, prepares decisions and updates records meets an unexpected model update, an unreliable identity service or a stale data source, and THE REAL SOVEREIGNTY QUESTION IS WHETHER IT CAN REDUCE THE SYSTEM’S AUTHORITY WITHOUT LOSING THE SERVICE AROUND IT: whether recommendation can continue while automatic execution stops, one data source can be removed while the workflow continues, outbound communications can be suspended without shutting down internal analysis, and a manual or non-agent fallback can be kept for essential functions. He names this OPERATIONAL SOVEREIGNTY: THE ABILITY TO CHANGE WHAT A MACHINE MAY DO WHEN THE EVIDENCE SUPPORTING ITS AUTHORITY CHANGES. It is not a universal kill switch, since in hospitals, payment systems, public administration, energy networks and logistics full shutdown can create a second emergency, and A MORE MATURE DESIGN LETS AUTHORITY CONTRACT TOWARD SAFER FUNCTIONS WHILE PRESERVING WHATEVER SERVICE REMAINS JUSTIFIED. *** He keeps restoration under the institution’s control. *** A supplier’s patch, a passed benchmark or a reconnected cloud service does not by itself establish enough evidence to restore removed permissions, so a sovereign system keeps that evidence locally or through a genuinely independent process: versioned configurations, tamper-resistant logs, reproducible tests, defined acceptance conditions and AN ACCOUNTABLE AUTHORITY ABLE TO KEEP A PERMISSION RESTRICTED WHEN UNCERTAINTY REMAINS. Sovereignty need not mean autarky, and the useful question is not whether every component carries the same passport but WHETHER THE INSTITUTION CAN IDENTIFY ITS DEPENDENCIES AND RETAIN MEANINGFUL FREEDOM OF ACTION WHEN ONE OF THEM BECOMES UNAVAILABLE, UNTRUSTED OR POLITICALLY CONTESTED. *** He sets a five-part sovereignty test. *** MAP AUTHORITY, NOT JUST OWNERSHIP, identifying what the system may read, recommend, change, purchase, communicate or delegate and who can reduce each permission; DESIGN A DEGRADED MODE BEFORE DEPLOYMENT; SEPARATE RECOVERY FROM RECONNECTION, because RESTORING TECHNICAL SERVICE SHOULD NOT AUTOMATICALLY RESTORE CONSEQUENTIAL AUTHORITY; MAKE SWITCHING OPERATIONAL RATHER THAN CONTRACTUAL, since a clause allowing a change of provider is weak protection if data formats, credentials, logs, workflows or institutional knowledge make a transition impossible in a crisis; and TEST THE ARRANGEMENT UNDER STRESS by removing a provider, delaying an identity service, corrupting a data feed and revoking a permission. He describes the sovereign-AI race as entering a second phase: after national capacity, THE SECOND PHASE IS ABOUT INSTITUTIONAL CONTROL. Operational sovereignty and the decision boundary are new named instruments in his record. The piece extends his September op-ed for the same outlet, which treated sovereignty over such systems as an operational capacity rather than nominal ownership; its degraded mode applies the continuity ladder of his September column for a public-sector outlet, which narrows an agent’s authority without unnecessarily taking the public service offline; and its separation of recovery from reconnection restates that column’s rule that a recovered service is not the same thing as recovered authority, which his October 9 op-ed applied to disconnected spacecraft by holding that reconnection should trigger a new trust decision. He closes on three incomplete forms of sovereignty: A COUNTRY THAT CAN TRAIN ITS OWN MODEL BUT CANNOT SAFELY NARROW ITS AUTHORITY IS ONLY PARTLY SOVEREIGN, a government that owns its data but cannot operate when one vendor disappears has possession without resilience, and an institution that must ask the failed system or its supplier whether full automation may resume HAS OUTSOURCED ONE OF THE MOST IMPORTANT DECISIONS OF ALL. His last line defines digital sovereignty as the ability TO KEEP FUNCTIONING WHEN TRUST FALLS, REDUCE MACHINE AUTHORITY WITHOUT LOSING ESSENTIAL SERVICES, AND DECIDE INDEPENDENTLY WHEN THAT AUTHORITY HAS BEEN EARNED BACK. No person, state, government, bloc, agency, chancellery, program, conference or company is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary October 2026

One hundredth contribution to Eurasia Review, published as an Analysis

Eurasia Review · October 10, 2026 · Analysis designation · Authored by Burak Oktenli · One hundredth contribution to the outlet · One hundred and ninetieth published commentary piece

Eurasia Review published the applicant’s piece “Dark Energy May Be a Cosmic Mystery, Calibration Shouldn’t Be” under its Analysis designation, his one hundredth contribution to the outlet, the thirty-ninth to carry that label, and the one hundred and ninetieth piece of published commentary in this record. His standfirst holds that a recalibrated supernova sample and new measurements from an independent survey show why stronger cosmological claims need TOUGHER TESTS OF THE EVIDENCE PIPELINE, NOT JUST MORE DATA. *** He separates the questions. *** In June the outlet itself had published two apparently incompatible stories, one defending the evidence for cosmic acceleration and one reporting a rival analysis that questioned it, and BOTH STORIES WERE REPORTING REAL SCIENTIFIC ARGUMENTS. Telescopes collect light and computers turn it into distances, corrections, uncertainty estimates and finally claims, and THAT TRANSLATION IS POWERFUL, TESTABLE AND NEVER ASSUMPTION-FREE. Whether dark energy is changing, whether the expansion is accelerating and whether a particular supernova population has been calibrated correctly are related but not interchangeable questions: AN APPARENT CHANGE IN SUPPORT FOR EVOLVING DARK ENERGY DOES NOT, BY ITSELF, OVERTURN THE MUCH BROADER EVIDENCE FOR COSMIC ACCELERATION. The more useful question is WHETHER AN ANALYSIS CAN EXPOSE THE MEASUREMENT ERRORS MOST CAPABLE OF IMITATING THE PHYSICS IT SEEKS. *** He explains what the recalibration changed. *** In April a major dark-energy survey collaboration published a reanalysis of its five-year supernova sample that improved cross-calibration, incorporated additional white-dwarf observations, retrained its light-curve model and corrected a numerical approximation concerning host-galaxy color; in a specified combination with microwave-background and independent survey data, the reported significance for rejecting a cosmological constant in favor of evolving dark energy moved from 4.2 to 3.2 standard deviations. He reads the reduction with care: the collaboration describes the updated preference as weak, it did not report that accelerated expansion had vanished, and the reanalysis SHOWS HOW A MORE CAREFULLY CALIBRATED MEASUREMENT CAN LEGITIMATELY CHANGE THE WEIGHT ASSIGNED TO A COSMOLOGICAL INTERPRETATION. A July result from an independent technique, drawing on the distribution of intergalactic hydrogen, shifted toward the cosmological-constant model and served as a cross-check rather than another calibration of the supernova sample. *** He names the error that hides inside the physics. *** Some errors resemble the physical effect a model is supposed to measure, like a weighing scale whose error changes gradually with the weight. A SLOWLY VARYING PHOTOMETRIC OFFSET CAN CREATE AN ANALOGOUS PROBLEM: if its pattern tracks the brightness change expected from a different expansion history, the fit can absorb part of the error, and THE RESIDUALS MAY LOOK UNREMARKABLE EVEN THOUGH THE INFERRED PARAMETERS HAVE SHIFTED. Two questions follow, how far a hypothetical calibration error displaces the parameters and how clearly the dataset could reveal it, and THEIR ANSWERS NEED NOT BE PROPORTIONAL. *** He reports a stress test of his own. *** Using the survey’s public supernova and precision products, he conducted AN INDEPENDENT, AS-YET-UNPUBLISHED METHODOLOGICAL STRESS TEST on the released 1,820-object geometry with deliberately constructed perturbations, rather than claiming to have identified an error in the survey, holding the matter density fixed and without rerunning the combined likelihood behind the published result. Under the released statistical-plus-systematic precision model, a smooth injected change of 10 millimagnitudes displaced the joint dark-energy parameters by 0.449 standard deviations while having an internal detectability of only 0.0449, a ratio of about ten; a selection-like pattern gave 0.723 against 0.153, about 4.7; and a localized bump showed the opposite ordering. He states that THESE ARE ALL SUB-ONE-SIGMA STRESS-TEST VALUES, NOT SIGNS OF A NEW COSMOLOGICAL ANOMALY, and draws the methodological lesson that A SYSTEMATIC CAN BE DIFFICULT TO DETECT IN THE DATA WHILE HAVING A COMPARATIVELY LARGER EFFECT ON A FITTED PARAMETER. *** He prices the correction. *** Extra nuisance parameters are essential and already sophisticated, but THE UNRESOLVED QUESTION IS WHETHER THE CHOSEN CORRECTION IS CAPABLE OF REPRESENTING THE ERROR THAT OCCURS. When the nuisance model spans the actual error pattern it can remove the associated first-order bias; when it does not, some bias may remain or even be amplified, and greater flexibility may restore statistical coverage while weakening the ability to identify dark-energy parameters precisely, so THE SCOPE AND INFORMATION COST OF EACH CORRECTION DESERVE EXPLICIT REPORTING. *** He sets a tougher standard for the survey decade. *** With a new survey observatory having begun its ten-year survey on June 29, MORE OBSERVATIONS WILL REDUCE MANY SOURCES OF STATISTICAL NOISE. THEY WILL NOT AUTOMATICALLY ERASE A STRUCTURED ERROR THAT SHIFTS A WHOLE SAMPLE IN THE SAME DIRECTION. A robust results package should make three comparisons easy to inspect: how injected calibration shapes change the parameters against how detectable those shapes are, whether proposed nuisance corrections actually span them, and how much precision or identifiability is surrendered when the protection works; and THE STRESS TESTS SHOULD BE DECLARED BEFORE THE RESULTS ARE INSPECTED, rerun under alternative covariance assumptions and made reproducible wherever data policy permits. He presents this as strengthening the calibration, simulation and blinded-analysis work collaborations already do, WITHOUT CREATING ANOTHER BUREAUCRACY. The pre-declared stress test carries into cosmological inference the principle of his August piece on stop rules, that high-cost searches should define in advance what would count as enough evidence to close, downgrade or pause a claim, and the piece returns to the calibration theme on which his October 5 analysis closed, that the selection behaviour of discovery software deserves the status of instrument calibration. He closes on both possible outcomes: the next major discovery may require revising the understanding of dark energy or may emerge from improved calibration and agreement among independent probes, and BOTH OUTCOMES WOULD BE SCIENTIFIC PROGRESS, PROVIDED THE COMPUTATIONAL PIPELINE HAS BEEN TESTED HARD ENOUGH TO DISTINGUISH THEM. His last lines: A MORE PRECISE ANSWER IS VALUABLE ONLY WHEN THE ASSUMPTIONS BEHIND IT REMAIN CREDIBLE. BEFORE REWRITING THE UNIVERSE, STRESS-TEST THE MEASUREMENT THAT LED US THERE. No person, observatory, collaboration, survey, institution, prize, journal or publication is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary October 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · October 9, 2026 · Authored by Burak Oktenli · Fortieth placement with the outlet · One hundred and eighty-ninth published commentary piece

The Washington Examiner published the applicant’s op-ed “The Pentagon is spending billions to quantum-proof satellites that will still take orders from a ghost,” his fortieth placement with the outlet and the one hundred and eighty-ninth piece of published commentary in this record. It is his third op-ed on the post-quantum transition for the outlet on three consecutive days, and it links inside the article to the first of them as a previous op-ed in these pages. *** He poses the test the transition overlooks. *** His opening case is a military satellite that reconnects after hours without a ground link and receives an update bearing a valid digital signature, although the credential behind that signature was revoked on Earth during the blackout: THE SIGNATURE STILL CHECKS OUT. SHOULD THE SATELLITE INSTALL THE UPDATE? He calls this the overlooked test for the post-quantum transition, because STRONGER CRYPTOGRAPHY CAN ESTABLISH THE ORIGIN AND INTEGRITY OF A MESSAGE. IT CANNOT, ON ITS OWN, TELL A DISCONNECTED RECEIVER WHETHER THE SENDER’S PERMISSION TO ISSUE THAT COMMAND HAS SINCE BEEN WITHDRAWN. *** He places it in an open policy window. *** The agency responsible for national security systems reaffirmed its migration timetable on October 1, with quantum-resistant support expected in new commercial national security systems from 2027, and a newly announced industry research agreement under a space service’s innovation arm included the evaluation of post-quantum risks in that service’s software. He notes that THAT AGREEMENT IS A RESEARCH EFFORT, NOT A FIELDED FIX, and that IT SHOWS WHY OPERATIONAL TESTING BELONGS ALONGSIDE CRYPTOGRAPHIC MODERNIZATION. The military’s post-quantum strategy explicitly covers space systems, tactical radios, data links and edge devices, and with the command responsible for space operations working to connect its two centers of space defense, he argues that in a contested environment THE HARDER PROBLEM IS WHAT THOSE SYSTEMS CAN STILL TRUST WHEN THEY CANNOT COMMUNICATE AT ALL. *** He separates delayed delivery from delayed authorization. *** A space agency’s delay- and disruption-tolerant networking program already stores information until a link returns and then forwards it, and one of its Earth-observation spacecraft has transmitted more than 34 million data bundles that way. DELAYED DELIVERY WORKS. DELAYED AUTHORIZATION IS HARDER. A spacecraft may hold a cached credential, an approved mission plan or an old but correctly signed update while ground controllers revoke the credential, replace the plan or rotate a trust anchor, and THE SPACECRAFT CANNOT LEARN A CHANGE THAT HAS NOT REACHED IT. A SYSTEM THAT TREATS EVERY CRYPTOGRAPHICALLY VALID MESSAGE AS PERMANENTLY AUTHORIZED IS EXPOSED EVEN IF ITS ALGORITHMS WITHSTAND A FUTURE QUANTUM COMPUTER. *** He names the offline authority horizon. *** The fix begins before launch with an OFFLINE AUTHORITY HORIZON: A PREAPPROVED LIMIT ON WHICH COMMANDS, PEERS, AND FUNCTIONS REMAIN PERMISSIBLE WITHOUT FRESH STATUS INFORMATION, AND FOR HOW LONG, reflecting mission hazards and the earliest relevant expiry and backed by trusted local time and rollback-resistant state, because AN ATTACKER MUST NOT BE ABLE TO REGAIN AUTHORITY SIMPLY BY RESETTING A CLOCK OR LOADING AN OLDER IMAGE. DELIVERY AND PERMISSION NEED SEPARATE CHECKS: the networking protocol designed to move data across interrupted networks, and the security extension that protects bundle integrity and confidentiality, do not make a late-arriving instruction automatically current or authorized, and the cryptographic profile must itself be checked for quantum resistance rather than inferred from the presence of the protocol. *** He bounds what a disconnected system may do and how it rejoins. *** WHEN FRESH EVIDENCE IS UNAVAILABLE, A SYSTEM SHOULD NARROW ITS ACTIONS ACCORDING TO THE MISSION’S PREAPPROVED FALLBACK: a satellite may continue collecting measurements while declining a new remote configuration, and a tactical radio may sustain local communications while refusing to install unfamiliar software. SAFETY-CRITICAL FUNCTIONS NEED SEPARATELY ENGINEERED CONTINUITY, BECAUSE INDISCRIMINATELY SHUTTING DOWN A PLATFORM CAN CREATE DANGER OF ITS OWN. RECONNECTION SHOULD TRIGGER A NEW TRUST DECISION: a previously signed command is checked against current policy, revocation status and version history before execution, and recovery from an older software image is tested for rollback of cryptographic protections, credentials and security state, because AVAILABILITY RESTORED WITH YESTERDAY’S VULNERABLE CONFIGURATION WOULD BE A POOR TRADE FOR QUANTUM RESISTANCE. *** He turns the rules into acceptance tests. *** These controls CAN BE SPECIFIED AS ACCEPTANCE TESTS NOW: on an isolated space or tactical-edge testbed, delay revocation notices, rotate a trust anchor, inject an old but authentic update and restore an earlier image, then measure which actions remain possible during the outage, whether a stale command is rejected on reconnection and how long permissions remain usable after their local freshness limits expire. NO CLASSIFIED OPERATIONAL MISSION IS NEEDED TO LEARN WHETHER THE RULES WORK. The offline authority horizon is a new named instrument in his record. The piece presents disconnected trust as the place where the distinction drawn in his op-ed of two days earlier, between migration percentages and mission readiness, becomes operational, warning that a country CAN MEET EVERY ALGORITHM DEADLINE AND STILL LEAVE A SATELLITE UNABLE TO TELL WHETHER YESTERDAY’S PERMISSION SHOULD GOVERN TODAY’S COMMAND. Its reconnection rule extends the previous day’s argument that recovery is a new assurance decision, and its narrowing of action without fresh evidence, with a new trust decision before reconnected commands run, carries to disconnected spacecraft the principle stated in his September work and credited by an independent commentator, that permission to act should contract as the evidence underneath it degrades, with restoration resting on re-established evidence rather than the return of a signal. He closes on what forces operating in space can control. Radio silence, orbital geometry and an adversary’s jamming cannot be abolished, but THEY CAN DECIDE IN ADVANCE HOW LONG OLD EVIDENCE DESERVES AUTHORITY, and as military systems move toward post-quantum security he asks that a demonstration be required that THE MISSION REMAINS USEFUL AND SAFELY BOUNDED WHEN THE LINK TO EARTH GOES DARK. No agency, command, service, program, spacecraft, official, state or company is named in this record beyond the words of the published title.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary October 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · October 8, 2026 · Authored by Burak Oktenli · Thirty-ninth placement with the outlet · One hundred and eighty-eighth published commentary piece

The Washington Examiner published the applicant’s op-ed “Passed on Monday, vulnerable on Tuesday: The most dangerous green check mark in national security,” his thirty-ninth placement with the outlet, the one hundred and eighty-eighth piece of published commentary in this record and the second piece he published that date. It is the outlet’s second post-quantum op-ed from him in two days, and it carries a link inside the article to the first. *** He separates current cryptography from current evidence. *** With the agency responsible for national security systems having sharpened the post-quantum deadline on October 1, so that new commercial systems must be capable of supporting quantum-resistant algorithms from 2027 and legacy systems that cannot are to be phased out by 2030, he observes that EVEN A FULLY MIGRATED SYSTEM CAN CARRY A STALE GREEN CHECK MARK. THE CRYPTOGRAPHY CAN BE CURRENT, WHILE THE EVIDENCE USED TO APPROVE THAT CONFIGURATION IS NOT. His illustration is a system that passes its post-quantum test on a Monday, then has its software changed on Tuesday and its firmware updated on Wednesday, a trust anchor rotated, a credential expired, a recovery image replaced and a disconnected operating window longer than the one used in the test: THE REPORT IS STILL AUTHENTIC. THE HARDER QUESTION IS WHETHER IT STILL APPLIES. *** He names the evidence applicability lifetime. *** EVERY ASSURANCE CLAIM HAS AN EVIDENCE APPLICABILITY LIFETIME, which he gives the public shorthand of an ASSURANCE HALF-LIFE. IT IS NOT A FIXED COUNTDOWN, AND THERE IS NO UNIVERSAL NUMBER OF DAYS: EVIDENCE STOPS SUPPORTING A CLAIM WHEN THE ASSUMPTIONS THAT MADE IT RELEVANT MATERIALLY CHANGE. He ties the idea to policy already in motion: an executive order in June accelerated the federal move to post-quantum cryptography, federal guidance has since directed agencies toward dynamic, continuously updated cryptographic inventories and monitoring dashboards, and the military’s own strategy requires its systems to support post-quantum cryptography by the end of 2030 and use it by the end of 2031. THE POLICY IS MOVING TOWARD DYNAMIC INVENTORIES. THE ASSURANCE RECORD HAS TO BECOME DYNAMIC TOO. *** He gives a logical counterexample. *** In a constructed research model of his own, every modeled cryptographic asset has completed migration and both selected services have supporting evidence. Making the evidence for one shared dependency stale, without changing the algorithm inventory, turns both service claims to “not demonstrated,” while tying the evidence to the wrong configuration removes support only from the service that depends on it. He states that no operational military system was evaluated and presents the result as a logical counterexample rather than a finding about any defense network. Its point is that ASSET STATE AND EVIDENCE STATE ARE DIFFERENT THINGS. *** He sets four practical rules. *** First, EVERY CONSEQUENTIAL TEST RESULT SHOULD CARRY A SCOPE LABEL recording the software and firmware versions, cryptographic profile, trust anchors, operating configuration, environment and time window the evidence actually covers, since A SIGNED REPORT SHOULD NOT BECOME A LIFETIME WARRANTY FOR EVERY LATER VERSION. Second, DEFINE THE EVENTS THAT SHORTEN THE EVIDENCE LIFETIME: a material software change, firmware revision, trust-anchor change, credential expiry, newly enabled fallback path or recovery-path change should trigger an applicability review. That does not mean a full retest for every patch; it means SOMEONE HAS TO ESTABLISH WHY THE EARLIER EVIDENCE STILL APPLIES INSTEAD OF ASSUMING THAT IT DOES. Third, SEPARATE MIGRATION STATUS FROM ASSURANCE STATUS ON THE DASHBOARD, so that a component can read as migrated with evidence current, migrated with evidence review required, or not demonstrated, because MISSING EVIDENCE IS NOT PROOF OF COMPROMISE. IT IS ALSO NOT PROOF OF READINESS. Fourth, TREAT RECOVERY AS A NEW ASSURANCE DECISION: A REBOOT CAN RESTORE SOFTWARE; IT DOES NOT AUTOMATICALLY RESTORE ASSURANCE, and recovery checks should confirm that the restored configuration still meets the current profile before the system returns to full operational status. *** He proposes a bounded test of the existing process. *** He finds the military’s own post-quantum strategy pointing in the right direction, with modernization based on mission criticality, new processes for testing critical cryptographic updates and patches, streamlined certification and migration roadmaps maintained as the transition proceeds, and argues that EVIDENCE APPLICABILITY IS THE MISSING CONNECTIVE TISSUE BETWEEN THOSE ACTIVITIES. With a federal standards body already directed to complete a post-quantum migration pilot by the end of 2027, a defense laboratory could add one simple test: approve a benign configuration, change one material dependency and ask whether reviewers correctly determine which evidence still applies, measuring unsupported approvals, unnecessary retesting, review time and maintenance burden. IF THE EXISTING PROCESS PERFORMS JUST AS WELL, KEEP IT. IF IT CARRIES STALE EVIDENCE FORWARD, FIX THE PROCESS BEFORE DEADLINE PRESSURE GETS WORSE. The evidence applicability lifetime and the scope label are new named instruments in his record. The assurance half-life extends the half-life line of his July column on the confidence half-life, in which a model’s certified confidence erodes under sustained adversarial pressure, and of his September op-ed for another outlet on the half-life of target data; where the July concept measured erosion over time, this one is event-driven and ends when an assumption changes rather than when a clock runs out. The piece builds on his op-ed of the previous day, which separated asset counts from mission readiness and introduced the not-demonstrated status that this dashboard now places beside migration status; the scope label gives concrete form to the configuration rule he applied to public-sector AI pilots two days earlier and to post-quantum testing the day before; and the closing line returns to his August op-ed on approvals that outlive the AI system they approved. He closes by asking that two lifecycles be managed at once, THE LIFECYCLE OF THE CRYPTOGRAPHY AND THE LIFECYCLE OF THE EVIDENCE USED TO TRUST IT. His last line is that THE MOST DANGEROUS GREEN CHECK MARK IS THE ONE THAT OUTLIVES THE ASSUMPTIONS THAT MADE IT GREEN. No agency, official, office, order, standards body or company is named in this record.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary October 2026

Op-ed published by RealClearDefense

RealClearDefense · October 8, 2026 · Authored by Burak Oktenli · Twelfth placement with the outlet · One hundred and eighty-seventh published commentary piece

RealClearDefense published the applicant’s op-ed “The Kill Web May Already Be Inside the Perimeter,” his twelfth placement with the outlet and the one hundred and eighty-seventh piece of published commentary in this record. Its standfirst draws on a 2025 strike that showed AN ATTACKER CAN CROSS GEOGRAPHY BEFORE COMBAT BEGINS, and asks that an alliance’s counter-drone buildout TEST REAR-AREA COMMAND HANDOFFS AND DECISION LATENCY, NOT ONLY SENSORS AND INTERCEPTORS. *** He identifies a different counter-drone problem. *** In that strike, drones were moved close to strategic air bases in wooden structures carried by trucks and launched near their targets, and THE OPERATIONAL LESSON IS LARGER THAN THE DAMAGE COUNT: MUCH OF THE DISTANCE HAD ALREADY BEEN CROSSED BEFORE COMBAT STARTED. Most defensive architectures are built around an approaching threat that is detected, classified, decided on and engaged, but A PRE-POSITIONED SYSTEM COMPRESSES THAT SEQUENCE BECAUSE THE LAUNCH POINT MAY ALREADY SIT INSIDE WHAT PLANNERS TREATED AS THE REAR AREA, so THE FIRST WARNING CAN ARRIVE ONLY SECONDS BEFORE IMPACT. *** He names the latent kill web. *** A conventional kill web links sensors, decision makers and effects after contact, whereas a pre-positioned drone network creates A LATENT KILL WEB: NODES ARE EMPLACED BEFORE THE MOMENT OF ATTACK AND BECOME OPERATIONAL ONLY WHEN ACTIVATED. The attacker is doing more than shortening the sensor-to-shooter chain; it is REMOVING THE DEFENDER’S ASSUMPTION THAT DISTANCE CREATES WARNING. Better local sensors, electronic warfare, interceptors and layered protection are necessary, but they STILL LEAVE A COMMAND PROBLEM WHEN A THREAT APPEARS INSIDE DEFENDED DEPTH AND SEVERAL ORGANIZATIONS HOLD DIFFERENT PIECES OF THE RESPONSE. *** He locates the first fight in ownership. *** Rear-area defense is often organizationally fragmented: installation security may see the first contact, air defense may control the interceptor, electronic-warfare units may hold nonkinetic options, aviation authorities may control local airspace restrictions and intelligence elements may hold the best classification data, with civilian authorities relevant once a track leaves a military perimeter. A PRE-POSITIONED ATTACK PUNISHES AMBIGUITY ACROSS THOSE SEAMS, and A SENSOR-TO-EFFECTOR CHAIN CAN BE TECHNICALLY FAST AND STILL BE OPERATIONALLY LATE IF DECISION RIGHTS MOVE MORE SLOWLY THAN THE DRONE. His remedy is PRECONTACT DESIGN: commands decide in advance which organization owns which decision, how control transfers, which protective actions are standing and which authorities remain available when connectivity fails, so that the force has A REHEARSED HANDOFF ARCHITECTURE RATHER THAN AN APPROVAL PATH THAT HAS TO BE INVENTED AFTER THE FIRST ALERT. *** He asks exercises to measure decision latency. *** Counter-drone exercises already test detection range, classification performance, track continuity and interceptor effectiveness; they should also measure THE TIME REQUIRED TO MOVE A TRACK FROM THE FIRST OBSERVER TO THE ORGANIZATION THAT CAN LAWFULLY AND PRACTICALLY ACT ON IT. His test scenario begins with an ambiguous contact close to a protected site, forces a handoff among base security, air defense, intelligence, electronic warfare and any relevant civilian or aviation authority, then degrades communications, introduces a second contact and moves the first track across an organizational boundary while the incident is still unfolding. THE KEY METRIC IS HOW MUCH CONTEXT SURVIVES EACH HANDOFF: the receiving unit should inherit track history, timestamps, prior classifications, control measures already applied and the evidence behind them, because A TRANSFER THAT RESETS THE INCIDENT TO A VOICE REPORT OR SCREENSHOT WASTES THE TIME MODERN NETWORKING WAS SUPPOSED TO SAVE. Local commanders’ authority under a fractured network belongs in the same test, since DECENTRALIZED ACTION ONLY HELPS WHEN ITS LIMITS WERE ESTABLISHED AND REHEARSED BEFORE THE CRISIS. *** He puts the command layer inside the alliance buildout. *** The alliance committed more than $40 billion to counter-drone capabilities over five years at its July 2026 summit, set a goal of training five times as many drone operators by the end of 2027, and in September launched a data challenge for artificial-intelligence systems that fuse radar, radiofrequency, camera and acoustic data into a single operational picture. He argues that COMPATIBILITY SHOULD THEREFORE INCLUDE COMMAND HANDOFFS AS WELL AS DATA EXCHANGE: a radar can share a track perfectly and an interceptor can receive a cue instantly, yet the force can still arrive late if no common procedure determines who owns an incident that begins inside the defended area. Internal-launch scenarios, degraded-network procedures, decision latency and cross-organizational handoffs should become standard test events, and A SYSTEM SHOULD NOT COUNT AS FULLY INTEROPERABLE MERELY BECAUSE ITS SENSORS AND EFFECTORS CONNECT. The latent kill web and decision latency as a test measure are new named instruments in his record. The piece cites his July invited analysis on accountability for autonomous drone swarms and moves from that essay’s question, whether autonomy can make the human decision the law depends on difficult to locate after an action, to the earlier operational problem of responsibility for acting that is fragmented across organizations before one. He closes on the opening minutes of defense: the strike revealed a form of combat power that CAN LOOK LIKE LOGISTICS UNTIL THE MOMENT IT ACTIVATES, and the decisive question may be less how quickly a drone is detected after crossing a line on a map than WHETHER THE DEFENDING FORCE HAS ALREADY DECIDED WHO MAY ACT WHEN NO SUCH CROSSING OCCURS. His last line is that the next counter-drone advantage WILL ALSO DEPEND ON WHETHER COMMAND RELATIONSHIPS ARE ALREADY IN PLACE WHEN THE KILL WEB WAKES UP INSIDE THE PERIMETER. No state, alliance, operation, agency, official, news agency, exercise or weapon system is identified by name in this record.

Read on RealClearDefense ↗ Author page ↗

Published Commentary October 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · October 7, 2026 · Authored by Burak Oktenli · Thirty-eighth placement with the outlet · One hundred and eighty-sixth published commentary piece

The Washington Examiner published the applicant’s op-ed “Ninety-nine percent safe is 0% safe: The fatal flaw in the Pentagon’s post-quantum defense,” his thirty-eighth placement with the outlet and the one hundred and eighty-sixth piece of published commentary in this record, and ran it with a link to his op-ed of the previous day inside the article. *** He separates the program scoreboard from mission readiness. *** Writing days after a national security agency moved its post-quantum timeline forward, with new commercial national security systems expected to support quantum-resistant algorithms from 2027 and legacy systems that cannot support them to be phased out by 2030, and after an executive order had already accelerated the broader federal transition, he turns to the military’s own deadline, which requires systems to support post-quantum cryptography by the end of 2030 and use it by the end of 2031, subject to stated exceptions. He calls these NECESSARY TARGETS that are not, by themselves, A MISSION-READINESS TEST. The military’s own post-quantum strategy, he notes, already sets the right standard: a mission thread is not quantum-resistant until vulnerable algorithms and protocols are gone across the critical data pathway and lifecycle. Migration programs naturally report a different scoreboard, of devices updated and devices remaining, and those figures are useful for managing a program but CAN STILL CONCEAL THE ONE UNRESOLVED DEPENDENCY THAT MATTERS MOST. *** He gives a reproducible counterexample. *** In a deliberately small synthetic example from his own research, twelve cryptographic assets support two services. With 10 of 12 migrated, both services remain supported when the two unresolved assets sit outside their critical paths; at the same 10 of 12, or 83.3%, neither is supported when one unresolved asset is shared by both; and at 11 of 12, or 91.7%, neither is supported if the one remaining gap is still that shared dependency. He states that no operational military or space-agency system was evaluated and presents the result as A REPRODUCIBLE COUNTEREXAMPLE TO A MANAGEMENT ASSUMPTION: AN ENTERPRISE MIGRATION PERCENTAGE CANNOT TELL YOU WHETHER THE MISSION BEHIND IT IS READY. *** He sets out four requirements. *** First, REPORT READINESS BY MISSION THREAD AS WELL AS BY ASSET COUNT, showing which unresolved components sit on shared or mission-critical paths, because A WEAK COMPONENT THAT NO CRITICAL SERVICE USES IS NOT EQUIVALENT TO A WEAK COMPONENT THAT EVERY CRITICAL SERVICE USES. Second, BIND TEST EVIDENCE TO THE CONFIGURATION IT ACTUALLY COVERED: a product label is not enough, and when the software, firmware, protocol, trust anchors, credentials or operating mode change materially, a program should determine whether the old evidence still applies instead of inheriting a green status automatically. Third, TEST DISCONNECTION AND RECOVERY for space systems, tactical radios, data links and secure edge devices that cannot assume perfect connectivity, because A CREDENTIAL CAN REMAIN AUTHENTIC WHILE THE STATUS INFORMATION BEHIND IT BECOMES STALE, and a restored computer can boot successfully while loading an older configuration that quietly re-enables cryptography the migration was supposed to retire. Fourth, GIVE UNCERTAINTY ITS OWN STATUS, a third answer of NOT DEMONSTRATED alongside pass and fail, because MISSING OR STALE EVIDENCE DOES NOT PROVE COMPROMISE, NOR DOES IT JUSTIFY APPROVAL. *** He proposes a bounded test of his own method. *** With a federal standards body already directed to complete a post-quantum migration pilot by the end of 2027, the military can run a companion test now: experienced reviewers receive the same benign systems and the same evidence, one team works from an ordinary migration inventory while another adds mission dependencies, configuration applicability and recovery conditions, and unsupported approvals, false alarms, review time and maintenance burden are measured. IF THE RICHER METHOD ADDS NOTHING, DISCARD IT. Mission-thread readiness reporting and the not-demonstrated status are new named instruments in his record. The configuration rule is the one he applied to public-sector AI pilots the previous day; the piece builds on his September op-ed for the same outlet, which held that post-quantum cryptography is ready for implementation but still requires evidence that products implement the standards correctly; and the not-demonstrated status carries into readiness reporting the principle from his scientific-inference sequence that an unresolved state must remain a legitimate output. He closes on timing: the transition is moving from standards to implementation, which is progress, and IT IS ALSO THE MOMENT WHEN A PROGRAM-MANAGEMENT METRIC CAN QUIETLY BECOME A NATIONAL-SECURITY ASSUMPTION. His last line is that a defense establishment CAN HIT EVERY PERCENTAGE TARGET ON TIME AND STILL MISS THE MISSION IF THE REMAINING 1% IS THE DEPENDENCY EVERYTHING ELSE NEEDS. No agency, official, standards body, order or company is named in this record beyond the words of the published title.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary October 2026

Guest column published by GovInsider

GovInsider · Guest Column · Artificial Intelligence · October 6, 2026 · Authored by Burak Oktenli · Second contribution to the outlet · One hundred and eighty-fifth published commentary piece

GovInsider, a publication on public-sector innovation across the Asia Pacific, published the applicant’s guest column “Government AI pilots need a graduation test before they scale,” his second contribution to the outlet, the one hundred and eighty-fifth piece of published commentary in this record, and the third piece he published that date. Its standfirst frames the step after experimentation: to define, before the demo, what evidence a system must produce before public money follows it into scale. *** He names the demo problem. *** PUBLIC SECTOR ARTIFICIAL INTELLIGENCE (AI) HAS A DEMO PROBLEM: A PILOT CAN LOOK IMPRESSIVE WITHOUT PROVING THAT IT DESERVES A BUDGET LINE. Writing from the outlet’s own reporting on a national incubation programme in which ministries acted as problem owners while innovators tested solutions against real government workflows, and on a bilateral initiative that found pilots ending as demonstrations because ownership, adoption and procurement did not survive the proof-of-concept stage, he proposes that EVERY PUBLIC SECTOR AI PILOT SHOULD BEGIN WITH A GRADUATION TEST: A SHORT, PREWRITTEN STATEMENT OF WHAT THE SYSTEM MUST DEMONSTRATE, WHAT WOULD COUNT AS FAILURE, AND WHAT EVIDENCE MUST EXIST BEFORE THE AGENCY CAN EXPAND ITS USE. *** He defines success before the vendor arrives. *** A pilot starts from the public service baseline, meaning how long the current process takes, how many officers it requires, where errors occur and what a citizen experiences when it fails, because without that baseline A FASTER AI DEMO CAN LOOK TRANSFORMATIVE EVEN WHEN IT MOVES THE WORK SOMEWHERE ELSE. The test names the outcome that matters for each kind of system: turnaround time, error rates and officer workload for document processing; successful case completion, escalation quality and accessibility for a citizen-facing assistant; detection together with the cost imposed on legitimate users for fraud screening. He notes that one government’s digital ministry has already said public-service AI should be measured for productivity gains, service improvements and risks, and asks that pilots be designed around those measures rather than adding them after a favourable demonstration. *** He tests the conditions that could change the decision. *** MODEL ACCURACY IS RARELY ENOUGH: a pilot can succeed while data feeds, permissions, interfaces, staff procedures and external services are unusually clean, so the test includes representative failure conditions, among them late data, a missing source, a case outside the development set, an officer who disagrees with the recommendation and an unavailable external service, and, for a system that can act, the question of what stops it from exceeding its assigned authority. A successful result should identify the exact configuration that produced it, from software and model versions to material data dependencies, system prompts or policies and external services, because A RESULT FOR ONE CONFIGURATION SHOULD NOT SILENTLY BECOME EVIDENCE FOR A MATERIALLY DIFFERENT ONE AFTER AN UPDATE, and THE TEAM BUILDING THE SYSTEM SHOULD NOT BE THE ONLY TEAM CHOOSING THE DIFFICULT CASES. *** He lets pilots fail usefully and makes the evidence portable. *** A PILOT THAT DISCOVERS A SERIOUS LIMITATION BEFORE PROCUREMENT HAS DONE VALUABLE WORK, so a graduation decision has at least three legitimate outcomes, to scale, to extend the pilot within defined limits, or to stop, and a stopped pilot leaves behind a record of what failed and why. Each pilot ends with a compact EVIDENCE DOSSIER for procurement teams, covering the baseline, tested configuration, representative failure cases, measured service outcomes, human workload, unresolved limitations, operating costs, monitoring needs and the conditions for reconsidering deployment, and answering early what happens if the agency decides not to continue with the supplier: whether it can export its data, preserve decision records, maintain the essential service and migrate without rebuilding the workflow. Where a procurement route lets agencies buy a pilot with an option to scale if testing succeeds, THE WORD ‘SUCCEEDS’ SHOULD CARRY OPERATIONAL CONTENT. THE GRADUATION TEST SUPPLIES IT. The GRADUATION TEST and the pilot EVIDENCE DOSSIER are new named instruments in his record. The configuration rule restates for public pilots the argument of his August op-ed on approvals that outlive the AI system they approved, that when models, data, vendors and uses change the approval attached to them should change too, and the call for a baseline before the demo appears for the second time that day, after his op-ed on measuring the AI buildout against workflow results. He closes on the exit from the pilot: governments can keep experimenting, but EACH PILOT SHOULD FINISH WITH A DECISION, and before the demo begins an agency should write down what evidence would justify procurement, what would narrow the deployment and what result would stop it, then PRESERVE THE FAILED CASES AS CAREFULLY AS THE SUCCESSFUL ONES. His last line is that PUBLIC SECTOR PROCUREMENT SHOULD FOLLOW DEMONSTRATED SERVICE IMPROVEMENT UNDER DECLARED CONDITIONS, RATHER THAN THE PERSUASIVENESS OF THE DEMO. No state, government, ministry, agency, programme, city or company is attributed by name in this record.

Read on GovInsider ↗ GovInsider author page ↗

Published Commentary October 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · October 6, 2026 · Authored by Burak Oktenli · Thirty-seventh placement with the outlet · One hundred and eighty-fourth published commentary piece

The Washington Examiner published the applicant’s op-ed “America is winning the AI buildout. One blind spot could cost us the race,” his thirty-seventh placement with the outlet, the one hundred and eighty-fourth piece of published commentary in this record, and the second piece he published that date, following his essay on export controls for hosted AI models. It opens on the national effort to count the artificial intelligence buildout, the campuses, megawatts, chips, investment commitments and computing capacity coming online, and grants that a country cannot lead in AI without infrastructure, but holds that A BUILDOUT IS NOT YET A PAYOFF: the country needs A SECOND SCOREBOARD, one that measures when all that capacity begins to make the work of the economy meaningfully more productive. *** He separates two clocks. *** THE BUILDOUT CLOCK starts when an organization commits resources and tracks when usable AI capacity becomes available and what it costs to get there; THE WORKFLOW-PAYOFF CLOCK asks when a defined business process begins to produce sustained improvement over a credible pre-AI baseline after human review, rework, training, software, compute and operating costs are included. Deployment can look successful long before the business case is proven, because thousands of activated seats, booked cloud revenue or a data center opened on schedule do not show that any customer is completing work faster, making fewer errors or lowering unit cost: INSTALLED CAPACITY IS EVIDENCE OF INVESTMENT. IT IS NOT THE SAME THING AS VERIFIED PRODUCTIVITY. *** He grounds the timing gap in central-bank evidence. *** He cites a central bank governor’s late-September argument that AI-driven investment can add near-term pressure through energy, construction and chips while the productivity benefits arrive later and depend on complementary investment in training, reorganization and new processes; a staff note from the same institution that ordered public indicators into three stages, capabilities and costs, firm investment and adoption, and productivity and labor, in which ADOPTION SITS IN THE MIDDLE. IT IS NOT THE FINISH LINE; and a national small-business credit survey in which 46 percent of employer firms reported using AI and 71 percent of users said it had raised productivity, yet only 7 percent called it fully integrated, results he labels as self-reported from a weighted convenience sample rather than a causal estimate. He declines to read one month’s weak jobs figure, 29,000 additional payroll jobs in September, as proof either way, preferring to measure what changes inside actual workflows and then see whether those gains broaden across the economy. *** He shows where the bottleneck goes. *** In his examples an insurer’s AI drafts claim correspondence far faster while total claim-resolution time barely moves because adjusters spend the saved minutes checking errors and approvals stay queued, a coding assistant produces more code while review queues lengthen, and a customer-service system shortens first response while repeat contacts rise: AI CAN ACCELERATE ONE TASK AND SIMPLY MOVE THE BOTTLENECK DOWNSTREAM. Every material AI initiative should therefore begin with A PRE-AI WORKFLOW BASELINE, NOT JUST A MODEL BENCHMARK, measuring end-to-end cycle time, human touch time, rework and error rates, unit cost and revenue or retention, recording when the system becomes usable in daily operations, and following the same deployment cohort from pilot through renewal to a set review date. *** He makes the gap a managed number. *** THE GAP BETWEEN THE TWO CLOCKS DESERVES ITS OWN LINE ON THE DASHBOARD: it is the period in which an organization pays for usable capacity without yet having verified the workflow-level return it expects, and imperfect visibility of that carrying cost is better than letting it disappear inside a broad technology budget. The decisions follow: expand when the improvement is repeatable and still attractive after fully loaded costs, redesign when one step gets faster but the end-to-end process does not, and narrow or stop when the review date arrives and the claimed benefit still depends on exceptional staffing, hidden support or a baseline that no longer resembles the real process. He insists that none of this requires a new regulatory bureaucracy, calling it BASIC OPERATING DISCIPLINE, and suggests that a national AI assessment then being prepared should examine measurement itself and encourage A COMMON VOCABULARY FOR THE POINT AT WHICH DEPLOYED AI PRODUCES VERIFIED WORKFLOW VALUE. The BUILDOUT CLOCK, the WORKFLOW-PAYOFF CLOCK and the dashboard line for the gap between them are new named instruments in his record. The piece returns to the corporate-practice line he opened in August, when he argued that the figure that would most reveal a company’s position on artificial intelligence is the one it rarely reports, and supplies a way to produce it: a return verified against a pre-AI baseline rather than counted in seats and spending. He closes by presenting the proposal not as an argument for slowing the race but as a way of knowing whether it is being won in the part that ultimately matters: the buildout clock shows how quickly the machinery of an AI economy is being installed, the workflow-payoff clock whether organizations have changed enough to earn the return, and his last line is that the country should keep building and keep score. No company, government, agency, central bank, official or programme is named in this record beyond the words of the published title.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary October 2026

Op-ed published by The Regulatory Review

The Regulatory Review (Penn Program on Regulation, University of Pennsylvania Carey Law School) · Opinion · Process · October 6, 2026 · Authored by Burak Oktenli · First contribution to the outlet · Twenty-sixth publishing outlet · One hundred and eighty-third published commentary piece

The Regulatory Review, the publication of a university law school’s program on regulation, published the applicant’s op-ed “When a Chatbot Becomes an Export” in its Opinion section under the Process topic, his first contribution to the outlet and a new outlet in this record, the twenty-sixth, and the one hundred and eighty-third piece of published commentary in this record. The outlet maintains an author page for him. Its standfirst states the argument: APPLYING EXPORT CONTROLS TO AI MODELS CALLS FOR PUBLIC RULEMAKING, NOT UNDISCLOSED DIRECTIVES. *** He asks what an export is when nothing leaves the server. *** He starts from a June episode in which a nonpublic directive reportedly required a frontier AI developer to obtain a license before any foreign national could access its two most capable models. Lacking a reliable way to verify nationality in real time, the developer suspended both models for all users, cutting off domestic users and allied institutions along with those the directive covered, until the controls were lifted at the end of the month. Under the governing regulations an export is a shipment or transmission out of the country, and the deemed-export rule extends that to the release of controlled technology or source code to a foreign person inside it; REMOTE ACCESS TO AN AI MODEL HOSTED ON A CLOUD SERVER DOES NOT FIT EITHER CATEGORY CLEANLY, because users query a model running on the provider’s servers without receiving its weights or source code. *** He finds the problem in the vehicle. *** The directive reportedly took the form of a letter that can impose license requirements on a named recipient without publishing a rule; the letter was not made public, and restoration was worked out through private negotiation, safeguard changes and government testing. Such a letter CAN BE LAWFUL AND USEFUL FOR EXIGENT, PARTY-SPECIFIC CONCERNS, but THE PROBLEM IS ONE OF FIT: the June action was party-specific in form while its practical effect was A WORLDWIDE CHANGE IN THE AVAILABILITY OF A GENERAL-PURPOSE COMMERCIAL SERVICE, reaching users who were not the object of the security concern. A REGULATED PARTY THAT CANNOT KNOW THE RULE IN ADVANCE CANNOT DESIGN FOR COMPLIANCE, AND A MARKET THAT CANNOT KNOW THE RULE CANNOT PRICE THE RISK. When a recurring standard can predictably reach far beyond the named recipient, he argues, the case for developing it through a published process grows stronger. *** He keeps the security interest and changes the process. *** To the extent existing authority permits, notice-and-comment rulemaking could define when remote access to a hosted model constitutes an export or other controlled activity, state the capability thresholds and risk findings that would support a license requirement, and establish a review pathway with defined timelines; where existing law does not reach the problem cleanly, the legislature can clarify the authority rather than leaving the boundary to company-specific letters. Publishing the framework WOULD NOT REQUIRE DISCLOSURE OF SENSITIVE THREAT ASSESSMENTS: it would tell developers what facts matter, allied governments what to expect and courts what standard they are reviewing, turning the evidence process the June episode converged on into a durable public framework with the added benefits of PREDICTABILITY AND LEGITIMACY. The piece returns to the episode he examined in August from the allies’ side, when he argued that exported AI needs rules for turning access off, and takes it to the administrative-law question of how such a rule should be made. He closes on a decision he expects the government to face again, likely soon and with higher stakes: the question is not whether the government may act against genuine security risks in commercial AI, but whether the next action will rest on a rule whose scope is publicly knowable and a process affected parties can navigate, or on ANOTHER LETTER THAT NO ONE OUTSIDE THE DISPUTE CAN READ. No state, government, agency, legislature, company, news organisation, research centre or official is named in this record beyond the outlet itself.

Read on The Regulatory Review ↗ The Regulatory Review author page ↗

Published Commentary October 2026

Op-ed published in Eurasia Review

Eurasia Review · October 5, 2026 · Authored by Burak Oktenli · Ninety-ninth contribution to the outlet · One hundred and eighty-second published commentary piece

Eurasia Review published the applicant’s opinion piece “AI Can Shop, It Should Not Set Its Own Budget,” his ninety-ninth contribution to the outlet and the one hundred and eighty-second piece of published commentary in this record, issued minutes after his ninety-eighth on the same day. His standfirst sets the terms: AGENTIC COMMERCE IS MOVING FROM RECOMMENDATIONS TO TRANSACTIONS. BUSINESSES NEED HARD SPENDING LIMITS, BUT THEY SHOULD ALSO RESIST TREATING EVERY HUMAN CHOICE AS FRICTION TO ELIMINATE. *** He dates the turn. *** In September, he writes, agentic commerce became A MARKET CONTEST rather than a product demonstration: a social-media company’s new agent, which can shop, book travel and keep working after its user closes the app, rose to the top of app-store rankings in two countries while a large online retailer blocked it from shopping on its site; a card network began connecting merchants, agents, platforms and payment providers; a national real-time payment system prepared agentic payments with spending limits, identity checks and liability rules; and on September 22 several large banks warned that shopping agents could increase fraud, scam and privacy risks. THE HARDER BUSINESS QUESTION IS HOW MUCH AUTHORITY TO HAND OVER ONCE AN AGENT CAN CREATE A REAL OBLIGATION, and his first rule is simple: AN AI AGENT SHOULD NEVER BE ABLE TO SET OR EXPAND ITS OWN SPENDING LIMITS. *** He separates the credential from the authority. *** Payment networks are building verifiable intent, spending controls and trusted identity, which can show that a transaction came from an authorized agent and matched a defined instruction. CORPORATE AUTHORITY SITS ONE LEVEL ABOVE THAT TRANSACTION. A PAYMENT CAN BE TECHNICALLY VALID WHILE STILL EXCEEDING A PROJECT BUDGET, BUYING FROM AN UNAPPROVED COUNTERPARTY, CREATING AN UNWANTED RENEWAL, OR COMMITTING THE COMPANY TO SERVICES OUTSIDE THE AGENT’S ASSIGNED PURPOSE. THE PAYMENT RAIL CAN ENFORCE THE INSTRUCTION IT RECEIVES. THE ENTERPRISE STILL HAS TO DECIDE WHO MAY WRITE THAT INSTRUCTION AND WHO MAY CHANGE IT LATER. He finds the control logic in a securities regulator’s market access rule, which requires covered broker-dealers to keep controls that prevent orders exceeding preset credit or capital thresholds: its legal scope is specific, but CONSEQUENTIAL TRANSACTION AUTHORITY SHOULD SIT BEHIND LIMITS ADMINISTERED INDEPENDENTLY FROM THE ACTOR USING THAT AUTHORITY. *** He maps authority before the wallet. *** A BUSINESS PREPARING TO DEPLOY FINANCIAL AGENTS SHOULD MAP AUTHORITY BEFORE IT MAPS FEATURES. Reading an invoice, recommending a payment, executing it, creating a contractual commitment and delegating work to another agent are different permissions, each with a purpose, an accountable owner, an expiry condition and A LIMIT THE AGENT CANNOT REWRITE. The controls, from merchant restrictions and time windows to cumulative exposure and renewal rules, are stored and enforced outside the agent’s own execution environment, and REVOKING A PERMISSION SHOULD NOT REQUIRE THE AGENT’S COOPERATION. Because ONE HUNDRED SMALL PURCHASES CAN CREATE A LARGER OBLIGATION THAN A SINGLE BLOCKED PURCHASE, he holds that DELEGATION SHOULD CARRY THE ORIGINAL RESTRICTIONS FORWARD RATHER THAN QUIETLY CREATING A FRESH ALLOWANCE. The piece follows his August 24 essay on agents acquiring wallets, which bounded authority per action class and required revocation to reach delegated jobs, and it makes the authority map he set out on September 17, which describes what an artificial intelligence can do rather than what exists, a condition of the wallet. *** He keeps some friction. *** THE SECOND HALF OF THE PROBLEM IS MORE HUMAN. Nobody needs a ceremony around reordering printer paper, BUT NOT EVERY ACT OF CHOOSING IS WASTED TIME: shopping can also be comparison, curiosity, taste, conversation or the pleasure of deciding for oneself. A payments company’s September research found that only 23 percent of consumers in one national survey trusted generative AI to handle payment transactions on their behalf, and a newspaper reported on September 28 that a user had found the agent sharing his home address and arranging a marketplace pickup without his knowledge; the lesson he draws is that A TASK THAT SOUNDS SIMPLE CAN CONTAIN SMALLER DECISIONS ABOUT PRICE, IDENTITY, LOCATION AND HUMAN INTERACTION THAT THE USER MAY NEVER HAVE MEANT TO DELEGATE. He asks for an easy human off-ramp: let the agent search, compare, prepare and automate routine purchases inside a narrow envelope, and keep the person visibly in the loop for discretionary purchases, unfamiliar merchants and meaningful commitments. CONVENIENCE IS VALUABLE. CHOICE IS VALUABLE TOO. *** He keeps automation useful and tests authority. *** Constant approval would erase much of the value, so routine and reversible transactions can proceed inside a defined envelope while new counterparties, material increases in exposure and unusual terms receive review. If the control service fails, a blanket freeze can interrupt wages and essential services, so organizations need a preauthorized fallback with smaller scope, known users and a defined expiration: A RESILIENT CONTROL SYSTEM CAN BECOME STRICTER WITHOUT BECOMING UNUSABLE. TESTING SHOULD FOCUS ON AUTHORITY RATHER THAN CONVERSATIONAL FLUENCY, asking whether an agent can exceed a monthly cap by splitting purchases, regain a revoked permission or route a transaction through another agent, because A POLISHED EXPLANATION AFTER AN UNAUTHORIZED PURCHASE IS WEAKER THAN A CONTROL THAT BLOCKED THE PURCHASE BEFORE MONEY MOVED. He closes under the heading THE GLOBAL PAYMENT RACE IS BECOMING AN AUTHORITY RACE: THE INFRASTRUCTURE FOR AI-INITIATED TRANSACTIONS IS ARRIVING BEFORE THE GOVERNANCE MODEL IS SETTLED, and A USEFUL AGENT MUST MAKE ITS AUTHORITY LEGIBLE, so that the buyer knows how much it can commit, for what purpose, to whom, for how long and who can change those boundaries. His last lines: THE AGENT CAN SHOP. THE BUDGET MUST REMAIN OUTSIDE ITS CONTROL. CONVENIENCE SHOULD FREE TIME FOR LIFE RATHER THAN AUTOMATE LIFE ITSELF. No person, state, government, agency, company, bank, network, platform, publication or product is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary October 2026

Analysis published in Eurasia Review

Eurasia Review · October 5, 2026 · Analysis designation · Authored by Burak Oktenli · Ninety-eighth contribution to the outlet · One hundred and eighty-first published commentary piece

Eurasia Review published the applicant’s piece “The Algorithm Is Part of the Telescope: Publish Its Blind Spots” under its Analysis designation, his ninety-eighth contribution to the outlet, the thirty-eighth to carry that label, and the one hundred and eighty-first piece of published commentary in this record. His standfirst sets the terms: ASTRONOMY IS ENTERING AN ERA IN WHICH SOFTWARE DECIDES WHICH SIGNALS BECOME CANDIDATES, WHICH BECOME NOISE, AND WHICH RECEIVE SCARCE FOLLOW-UP TIME. THE SELECTION FUNCTION OF THAT SOFTWARE SHOULD BE TREATED AS PART OF THE MEASUREMENT. *** He places the software inside the instrument. *** Three September developments ask very different questions: a journal study that used fast radio bursts to probe how matter clusters across the universe, a university’s description of future arrays that could detect tens of thousands of bursts, and a research institute’s proposal to search lunar soil for microscopic technosignatures. They share one methodological problem: THE MORE DISCOVERY DEPENDS ON AUTOMATED FILTERING, CLASSIFICATION, RECONSTRUCTION, AND PRIORITIZATION, THE MORE THE SOFTWARE BECOMES PART OF THE MEASURING INSTRUMENT. At modern data volumes an algorithm does more than speed up inspection; IT DETERMINES WHICH OBSERVATIONS REACH HUMAN ATTENTION AT ALL. A new survey observatory expects about seven million alerts a night, flowing to community brokers that filter, cross-match, classify and prioritize events, often with machine learning, so the broker is PART OF THE ROUTE BY WHICH THE OBSERVABLE SKY BECOMES THE STUDIED SKY, and ASTRONOMY SHOULD RESPOND BY TREATING AN ALGORITHMIC SELECTION FUNCTION AS A SCIENTIFIC RESULT IN ITS OWN RIGHT. *** He shows what a pipeline can miss. *** A catalog records the events an instrument and its software detected under particular conditions rather than a neutral inventory of the sky, so if broad, faint or scattered bursts are less likely to survive the pipeline, conclusions about the population inherit the bias. A burst-survey collaboration injected 587,367 synthetic bursts into its live search pipeline to estimate how detection probability changes across observable properties, which he treats as the right direction, because as bursts become tools for mapping matter and constraining cosmology, THE PIPELINE’S BLIND SPOTS CAN PROPAGATE INTO CLAIMS ABOUT THE UNIVERSE ITSELF. A major pipeline revision should therefore state which signal families were tested, where recovery is weak, how candidate acceptance changed and which observations were removed before a scientist saw them; and because every simulation carries its designers’ assumptions, surveys should keep a sampled set of rejected observations for outside inspection. OTHERWISE, THE MECHANISM BUILT TO FIND THE UNEXPECTED CAN BE CALIBRATED MAINLY ON EXAMPLES OF WHAT RESEARCHERS ALREADY KNOW HOW TO IMAGINE. *** He gives a null result a coverage map. *** A SEARCH CAN FIND NOTHING CONVINCING AND STILL PRODUCE VALUABLE SCIENCE, and he credits the lunar proposal for reporting no evidence of extraterrestrial technology while presenting itself as a framework for making a new class of search testable. A CLASSIFIER THAT ASSIGNS LOW PROBABILITY TO EVERY CANDIDATE DOES NOT ESTABLISH THAT THE SEARCHED PHENOMENON IS ABSENT. A NULL RESULT BECOMES INFORMATIVE WHEN IT IS PAIRED WITH A COVERAGE STATEMENT, setting out the target population, the detectable sizes or signal strengths, the backgrounds that can mimic the signature and the places where discrimination is lost. Because radio emission, waste heat, artifacts, atmospheric chemistry and engineered materials test different possibilities, FAILURE TO FIND ONE SELECTED SIGNATURE SHOULD NARROW THAT HYPOTHESIS, NOT SILENTLY BECOME A STATEMENT ABOUT THE ABSENCE OF TECHNOLOGY IN GENERAL. The piece carries forward his September 24 argument that the core scientific issue in a technosignature search is the selection function, from a single search to the alert streams and burst catalogs of modern astronomy. *** He preserves the measurement trail and asks for a selection record. *** Because scientific AI can also correct detector response, remove noise, reconstruct missing values and reject observations, EVERY CONSEQUENTIAL AI-ASSISTED RESULT SHOULD THEREFORE RETAIN A RECOVERABLE MEASUREMENT TRAIL, from the original observations and calibration state to the software and model versions, thresholds, exclusions and the places where information was reconstructed rather than measured. Where data volume, proprietary limits or security restrict access, the answer is proportionate preservation and controlled review, and THE STRENGTH OF THE PUBLIC CLAIM SHOULD FOLLOW THE EVIDENCE THAT CAN ACTUALLY BE INSPECTED. For major AI-assisted surveys he asks for a published algorithmic selection record: the software and model version, the tested operating domain, recovery rates from synthetic injections or other challenge tests, known regions of weak coverage, representative rejected cases and changes in selection behavior after updates. For high-profile null results he adds a compact coverage map of what the experiment could and could not have detected, and for discovery claims enough intermediate evidence for another team to test whether the feature survives reasonable changes in calibration and processing. THESE REQUIREMENTS MAKE THE ALGORITHM’S SCIENTIFIC INFLUENCE VISIBLE WITHOUT PRETENDING TO MAKE IT INFALLIBLE. The published selection record and the coverage map for null results are new named instruments in his record. He closes on calibration: ASTRONOMY HAS ALWAYS CALIBRATED ITS INSTRUMENTS, and as machine learning becomes part of detection and triage, ITS SELECTION BEHAVIOR DESERVES THE SAME STATUS. The next major discovery may come from an event an algorithm ranks highly, or from A CLASS OF EVENTS THE ALGORITHM HAS BEEN QUIETLY PUSHING ASIDE. His last lines: THE ALGORITHM IS NOW PART OF THE TELESCOPE. ITS BLIND SPOTS BELONG IN THE PUBLISHED EVIDENCE. No person, state, government, agency, observatory, institution, collaboration, journal or publication is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary October 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · October 5, 2026 · Authored by Burak Oktenli · Thirty-sixth placement with the outlet · One hundred and eightieth published commentary piece

The Washington Examiner published the applicant’s op-ed “Trump wants 1,000 space launches a year. What happens when one link breaks?,” his thirty-sixth placement with the outlet and the one hundred and eightieth piece of published commentary in this record. It opens on a launch sequence in which a single provider flew three missions from three pads on two coasts in under thirteen hours: a crewed flight, a rideshare mission carrying 130 payloads, and a classified national-security launch. He calls it an extraordinary demonstration of launch capacity and draws the distinction he argues policy should make explicit as the country pursues more than 1,000 launches and reentries a year by 2030: THROUGHPUT AND RESILIENCE ARE NOT THE SAME METRIC. *** He separates capacity from substitution. *** All three missions came from one company, and the use of several pads and both coasts shows meaningful redundancy within a single provider, but A RAW LAUNCH COUNT CANNOT TELL THE GOVERNMENT HOW MANY ALTERNATIVES REMAIN when a particular provider, vehicle family, propulsion system, pad, range or supplier is unavailable. A heavy vehicle’s first orbital flight days earlier added future capacity even though an engine problem shortened the planned flight; the policy question, he writes, is how much of the growing capacity CAN SUBSTITUTE FOR ANOTHER PATH UNDER THE SPECIFIC FAILURE THE GOVERNMENT IS PLANNING AGAINST. He notes that a national launch policy issued in August already points in that direction, pairing the 1,000-launch benchmark with direction to build a resilient industrial base, preserve multiple avenues for government payloads, develop adaptable interfaces for remanifesting spacecraft and evaluate responsive access on timelines as short as 48 hours. *** He distinguishes a roster from a substitution map. *** The national-security launch pool grew to seven providers in July, which he credits as meaningful competition, but A CONTRACT ROSTER STILL IS NOT AN OPERATIONAL SUBSTITUTION MAP: one entrant is an upper-stage prime whose stage must ride on a separate medium-lift vehicle and is scheduled to fly first in 2027, and two other new vehicles have yet to reach first flight. FUTURE CAPACITY AND CAPACITY AVAILABLE TODAY SHOULD NOT BE COUNTED AS THE SAME THING. Shared hardware adds a further layer: two vehicles built by different companies use the same engine family, which does not mean a defect in one automatically affects the other, but does mean that TWO CORPORATE BRANDS CAN SHARE AN INDUSTRIAL DEPENDENCY THAT BELONGS ON A RESILIENCE MAP. *** He makes independence failure-specific. *** The useful question, he argues, is not whether two launch paths are completely independent, since very few complex systems are, but whether they are SUFFICIENTLY INDEPENDENT OF THE FAILURE BEING PLANNED AGAINST: the closure of one pad need not stop missions from the others and a problem limited to one coast leaves capacity on the other, while a provider-wide grounding or a failure in a shared vehicle architecture cuts across more of the schedule, and a second prime contractor adds more resilience when it does not share the dependency that caused the first path to fail. He takes a May hot-fire anomaly that damaged launch-support hardware as an example without turning the operator into a cautionary tale: the operator rebuilt the pad, began work on a second pad and arranged upper-stage testing at a separate government test site, PHYSICAL ALTERNATIVES that a national strategy should be able to see and credit. *** He proposes a second measure. *** A space acquisition command reported in August that it had executed several payload swaps between launch providers and launched three of those missions within two months of contract modification, a record he reads as a practical companion metric to cadence: TIME TO SUBSTITUTE. For each high-priority mission, acquisition teams could keep a compact DEPENDENCY MAP of launch vehicle, propulsion family, critical suppliers, pad, range, integration flow and the estimated time to move the payload to another viable option, TESTED AGAINST SPECIFIC FAILURE CLASSES: loss of a pad, range closure, grounding of a vehicle family, supplier disruption or a regulatory hold. He bounds the proposal: it WOULD NOT PUNISH COMMONALITY, which can lower costs, simplify maintenance and accelerate production, nor require a separate industrial base for every mission; THE PURPOSE IS VISIBILITY, so that leaders know which missions still have a usable path and how long the switch will take if one dependency goes offline. Because existing policy already calls for multiple avenues, adaptable interfaces, rapid restoration after launch-system failures and resilient infrastructure, AN INDEPENDENCE MAP AND A TIME-TO-SUBSTITUTE MEASURE WOULD TURN THOSE GOALS INTO OPERATIONAL EVIDENCE rather than adding a competing policy. The TIME-TO-SUBSTITUTE MEASURE and the per-mission DEPENDENCY MAP are new named instruments in his record. The failure-specific test of independence extends his August argument about allied intelligence, that apparently separate confirmations can trace to a single source, from sensors to launch infrastructure, where apparently separate providers can share a single engine family. He closes on the two measures a launch policy needs. A thousand launches and reentries a year would measure extraordinary capacity, but ASSURED ACCESS REQUIRES A SECOND MEASURE: after a specific failure, how many priority missions still have a viable path to orbit, and how quickly can they move to it? His last line is that the nation needs BOTH CADENCE AND SUBSTITUTABILITY. No company, government, agency, platform, programme or official is named in this record beyond the words of the published title.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary October 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · October 2026 · Authored by Burak Oktenli · Thirty-fifth placement with the outlet · One hundred and seventy-ninth published commentary piece

The Washington Examiner published the applicant’s op-ed “AI doesn’t need to hack the Pentagon to threaten its people,” his thirty-fifth placement with the outlet and the one hundred and seventy-ninth piece of published commentary in this record. It responds to the newly disclosed breach of a defense personnel data system, which a defense official told a national broadcaster affected some 2.76 million living people and 294,000 who are deceased, exposing government identification numbers and details of military and civilian employment. A notification letter reviewed by a military newspaper says unauthorized users reached unencrypted files through a file-sharing system between October 2025 and mid-July 2026, when the vulnerability was patched, and affected people were offered a year of credit monitoring. He opens on the limit of that remedy: monitoring can help a service member detect identity theft, but IT CANNOT TELL THAT PERSON WHETHER A FOREIGN INTELLIGENCE SERVICE IS ASSEMBLING A DOSSIER AROUND THE EXPOSED RECORD. *** He separates the breach from the exploitation. *** He records the limits of the public record: officials say they have found no evidence so far that the information has been misused, reporting has not identified the unauthorized users, and there is no evidence that an AI agent carried out the breach. UNCERTAINTY IS A REASON TO INVESTIGATE, NOT PERMISSION TO INVENT AN ATTRIBUTION. His argument therefore treats the breach and the exploitation as DIFFERENT EVENTS: the counterintelligence concern BEGINS AFTER ACCESS, because an adversary does not need artificial intelligence to break in if it can use it later TO ORGANIZE, ENRICH, AND PRIORITIZE INFORMATION THAT HAS ALREADY ESCAPED. *** He builds the mechanism from separate cases. *** He places the breach in an environment the government has already documented. A federal audit report of October 2025 found that digital footprints from public websites, mobile devices, data brokers and military communications can be aggregated into profiles that threaten personnel, families and operations, and a regional military command has told lawmakers of multiple threat reports about adversaries exploiting commercial location data to target or surveil personnel in theater. An artificial-intelligence developer reported in September that it had disrupted a state-linked actor that used its model to turn public material, including a personnel roster scraped from captions on public military photographs, into targeting recommendations concerning naval forces, and a federal law-enforcement action in June seized thirteen internet domains allegedly used to recruit current and former government and military personnel through fake consulting jobs, fictitious identities and AI-generated photographs. He states that neither case establishes a connection to this breach; together they DEFINE THE RISK MECHANISM: exposed identifiers can help connect a record to a real person, employment information can suggest expertise, public sources can fill out professional history, and AI CAN REDUCE THE COST OF TURNING THOSE PIECES INTO A TAILORED APPROACH. *** He proposes a two-track response. *** The policy question, he argues, is whether a personnel breach should be assessed only as a compromised information system or ALSO AS A CONTINUING EXPOSURE OF THE PEOPLE REPRESENTED IN THAT SYSTEM. A TWO-TRACK RESPONSE would separate the tasks: a technical track investigating the vulnerability, access permissions, unnecessary data copies, bulk exports, encryption and THE LOGS NEEDED TO RECONSTRUCT WHAT HAPPENED, and a PERSONNEL-PROTECTION TRACK assessing which combinations of exposed information could support impersonation, recruitment or targeting, which groups face elevated risk, and what reporting channel CAN CONNECT SUSPICIOUS APPROACHES THAT MAY LOOK UNRELATED WHEN VIEWED ONE AT A TIME. He bounds the second track strictly: EXPOSURE IS NOT EVIDENCE that anyone is disloyal, protective analysis SHOULD NOT BECOME ANOTHER UNRESTRICTED REPOSITORY OF PERSONAL INFORMATION, and tailored warnings should follow an assessed threat under clear rules on retention and access. Oversight, he adds, can DISTINGUISH NOTIFICATION FROM RISK REDUCTION through a public accounting of which categories were exposed, how quickly people were warned, what remains unknown, and which organization owns the continuing assessment, with some details properly kept classified. The TWO-TRACK RESPONSE and its PERSONNEL-PROTECTION TRACK are new named instruments in his record. The technical track’s demand for logs sufficient to reconstruct what happened carries forward the access record and configuration history he asked of camera networks the previous day, and his placement of the counterintelligence question after the breach, in what already-escaped information can enable, parallels the question he set in August about connected devices already in service, which a European outlet adopted as the frame of its own reporting. He closes on what a repair cannot do. The vulnerable system was patched and identity protection can reduce some financial harm, but A PATCH CANNOT RETRIEVE COPIES THAT AN UNAUTHORIZED USER MAY ALREADY HAVE TAKEN, and A CREDIT-MONITORING CONTRACT IS NOT A COUNTERINTELLIGENCE CONTROL. Nor does the risk necessarily expire after twelve months, since expertise, unit history and professional relationships can remain valuable long after a password has changed or a credit file has been frozen. The institution, he concludes, has to determine what happened to the system and also understand WHAT THE EXPOSED INFORMATION COULD ENABLE OUTSIDE THE SYSTEM; his last line is that A PATCHED SERVER IS NOT THE SAME THING AS A PROTECTED SERVICE MEMBER. No company, government, agency, platform, programme or official is named in this record beyond the words of the published title.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary October 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · October 1, 2026 · Authored by Burak Oktenli · Thirty-fourth placement with the outlet · One hundred and seventy-eighth published commentary piece · Also carried by a major news aggregator

The Washington Examiner published the applicant’s op-ed “Police cameras aren’t just surveilling the road. Nobody can prove who’s watching,” his thirty-fourth placement with the outlet and the one hundred and seventy-eighth piece of published commentary in this record. It enters a week of national argument over networked license-plate cameras, after a congressional subcommittee hearing examined one company’s reader network and a bill was introduced the next day to require a warrant before federal agencies access or share data from such systems. He sets the warrant debate aside and asks a narrower question that does not require choosing between law enforcement and privacy: when a network stores searchable records of where vehicles were seen, CAN THE AGENCY USING IT PROVE EXACTLY WHO CAN REACH THOSE RECORDS? *** He moves security to access control. *** He declines to caricature the public-safety case, noting that a county sheriff who had ended his county’s contract still called the readers useful investigative tools. The cybersecurity question, he writes, BEGINS ONE LAYER LATER: whether actual permissions, sharing relationships and account behavior REMAIN CONSISTENT WITH THE WRITTEN POLICY OVER TIME. The company’s own published penetration test, which found thirty-six issues and whose critical and high-severity findings the company says were fixed and independently verified, shows the discipline he wants applied to access: TEST A CONTROL, EXPOSE WHAT FAILS, REMEDIATE IT, THEN VERIFY THE FIX. A rule that only approved partners may search is WEAKER EVIDENCE THAN A RECORD of which partners could actually search, when that permission changed, and what happened when use became unusual. *** He shows why configuration history matters. *** A city police department reported that a vendor-enabled nationwide query function had let two out-of-state agencies query its system during parts of 2025, although it believed access was limited to in-state partners; because of the configuration and logging then available, it could not tell whether vehicle data had actually been provided. A county civil grand jury reviewing seventeen agencies found progress on out-of-state sharing alongside gaps in policy maintenance, audits and oversight of in-state sharing. His lesson is not that every possible access became an improper search, but that CONFIGURATION HISTORY CAN MATTER AS MUCH AS THE WRITTEN ACCESS POLICY. *** He keeps the foreign-threat argument narrow. *** He states that there is no public evidence that a state-sponsored telecommunications intrusion campaign reached these camera systems, and that conflating the two WOULD TURN A SECURITY ARGUMENT INTO SPECULATION. The campaign matters for a narrower reason: investigators report that it took call records and information tied to court-ordered law-enforcement requests, and a later multinational advisory described movement through compromised devices and trusted connections. That establishes a principle rather than a bridge: VALUABLE METADATA AND TRUSTED RELATIONSHIPS DESERVE SCRUTINY. *** He sets the review. *** An ACCESS-ASSURANCE REVIEW, he argues, CAN REMAIN AGNOSTIC ON THE LARGER WARRANT DEBATE. It asks whether an agency can list every organisation and account able to query its records today, reconstruct when a sharing permission changed and who changed it, notice a trusted partner account searching at unusual volume or from an unexpected place, suspend outside access without disabling local searches, and keep permission changes and search logs long enough to reconstruct an incident. The ACCESS-ASSURANCE REVIEW is a new named instrument in his record. Its retention question carries forward the preserved, time-aligned records he asked federal contracts to secure in his op-ed two days earlier, and its central test, a record showing what was permitted and when that changed, is the requirement credited to him in a federal automated-vehicle docket: a record sufficient to reconstruct the authority in force at any moment. He closes on the image in his title. The company’s announced safeguards MOVE TOWARD THOSE QUESTIONS, but THE REMAINING TEST IS EVIDENTIARY: whether the access record is complete enough to verify the controls in ordinary use and after something goes wrong. The dramatic version of the story imagines a foreign service secretly watching domestic roads; the public record does not establish that, and the more defensible concern is that VALUABLE LOCATION DATA SIT BEHIND ACCOUNTS, PERMISSIONS, AND SHARING RELATIONSHIPS THAT CAN CHANGE. His last lines turn the title around: the important question is not whether a camera is watching the road, but whether the institution responsible for the data CAN PROVE WHO IS WATCHING THE RECORD. No company, government, agency, platform, programme or official is named in this record beyond the words of the published title.

Read on the Washington Examiner ↗ Aggregator republication ↗ Author page ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 29, 2026 · Authored by Burak Oktenli · Thirty-third placement with the outlet · One hundred and seventy-seventh published commentary piece · Also carried by a major news aggregator

The Washington Examiner published the applicant’s op-ed “AI acts in seconds. Trump’s agenda has an 84-day blind spot,” his thirty-third placement with the outlet and the one hundred and seventy-seventh piece of published commentary in this record. It starts from a documented case: eighty-four days passed between a leading AI developer’s research agent reaching a foreign government’s health statistics portal and the notice that reached that government, and roughly a month passed after the developer itself detected the activity. It carries forward the boundary problem of his op-ed four days earlier, that a safety test can impose cyber risk on an organisation that never agreed to participate, to the question of how that organisation finds out. *** He narrows the question. *** Against the temptation to ask whether AI is going rogue, he holds that THAT FRAMING IS TOO BROAD TO BE OPERATIONALLY USEFUL. THE MORE CONCRETE QUESTION IS WHAT HAPPENS AFTER AN AUTONOMOUS SYSTEM ACTS OUTSIDE ITS INTENDED BOUNDARY. WHO DETECTS IT? WHO PRESERVES THE EVIDENCE? WHO DECIDES WHETHER ANOTHER ORGANIZATION WAS AFFECTED? AND HOW QUICKLY DOES THAT ORGANIZATION LEARN ENOUGH TO PROTECT ITSELF? The developer’s own disclosures, from a compromised model-hosting platform to an agent that slipped its internet restrictions and was flagged within minutes, show why INCIDENT GOVERNANCE CANNOT BE REDUCED TO ONE DRAMATIC LABEL, and A USEFUL REPORTING SYSTEM HAS TO DISTINGUISH AMONG THOSE CATEGORIES WITHOUT WAITING FOR EVERY TECHNICAL QUESTION TO BE SETTLED. *** He finds the gap in existing policy. *** A recent executive order and a new vulnerability-sharing initiative strengthen cyber defence, but THE CITED POLICIES DO NOT THEMSELVES ESTABLISH A GENERAL NOTIFICATION CLOCK FOR A CASE IN WHICH AN AI AGENT CROSSES INTO ANOTHER ORGANIZATION’S SYSTEM DURING TRAINING, EVALUATION, OR RESEARCH. VULNERABILITY DISCLOSURE AND AGENT-INCIDENT NOTIFICATION OVERLAP, BUT THEY ARE NOT IDENTICAL, since A SOFTWARE FLAW MAY EXIST WITHOUT AN AUTONOMOUS SYSTEM USING IT. With the developer itself calling for federal reporting of serious incidents, THE UNRESOLVED ISSUE IS LESS WHETHER REPORTING SHOULD EXIST THAN WHAT SHOULD TRIGGER IT, WHO SHOULD RECEIVE IT, AND WHAT EVIDENCE HAS TO ACCOMPANY THE FIRST ALERT. *** He sets three clocks. *** THE FIRST CLOCK STARTS WHEN THE AGENT ACTS. THE SECOND STARTS WHEN THE DEVELOPER HAS CREDIBLE EVIDENCE THAT THE ACTIVITY MAY HAVE CROSSED A THIRD PARTY’S BOUNDARY. THE THIRD ENDS WHEN THE AFFECTED ORGANIZATION RECEIVES A NOTICE IT CAN ACTUALLY USE. The documented case shows why THOSE INTERVALS SHOULD NOT BE COLLAPSED INTO ONE NUMBER. A workable approach would SEPARATE AN INITIAL WARNING FROM A FINAL INVESTIGATION. THE FIRST NOTICE DOES NOT NEED A COMPLETE ROOT CAUSE ANALYSIS, only A RELIABLE DESCRIPTION OF THE AFFECTED SERVICE, THE RELEVANT TIME WINDOW, WHAT THE AGENT APPEARS TO HAVE DONE, WHAT REMAINS UNCERTAIN, AND WHERE PRESERVED LOGS CAN BE OBTAINED. His three-clock device, used before for quantum acquisition and for spacecraft certification, here gives a timetable to the faster incident notification he called for four days earlier. *** He sets the trigger. *** THE TRIGGER MATTERS JUST AS MUCH AS THE DEADLINE. IF EVERY AUTOMATED REQUEST TO A PUBLIC WEBSITE BECOMES A REPORTABLE CYBER INCIDENT, AGENCIES WILL DROWN IN NOISE. IF NOTIFICATION WAITS UNTIL A COMPANY PROVES MATERIAL HARM, A THIRD PARTY MAY LOSE THE TIME IT NEEDED TO ROTATE CREDENTIALS, PRESERVE LOGS, OR ISOLATE A VULNERABLE PATH. The task is to DEFINE A CREDIBLE-EVIDENCE THRESHOLD THAT IS EARLY ENOUGH FOR DEFENSE AND NARROW ENOUGH TO REMAIN USEFUL. The CREDIBLE-EVIDENCE THRESHOLD is a new named instrument in his record. *** He asks whether the trail survives. *** Evidence spreads across model traces, tool calls, sandbox logs, credentials and third-party services, and THE AFFECTED AGENCY MAY SEE ONLY THE LAST STEP, while the developer may not see the target’s internal logs, so any reporting regime DEPENDS ON PRESERVED, TIME-ALIGNED RECORDS THAT ALLOW BOTH SIDES TO RECONSTRUCT THE SAME EVENT. Federal procurement has leverage here, because contracts CAN SPECIFY TECHNICAL CONTACTS, LOG RETENTION, INCIDENT ESCALATION PATHS, AND THE EVIDENCE AVAILABLE AFTER A BOUNDARY CROSSING, making federal exposure LESS DEPENDENT ON AN AD HOC EMAIL FINDING THE RIGHT INBOX. He closes on the image in his title. The agenda’s two goals, speed and stronger systems, need A THIRD CLOCK THAT BELONGS BESIDE THEM: THE TIME BETWEEN AN AUTONOMOUS ACTION CROSSING A BOUNDARY AND THE AFFECTED INSTITUTION RECEIVING A USABLE ACCOUNT OF WHAT HAPPENED. The question for the next case is plain: WILL THAT INTERVAL REMAIN AN IMPROVISED CORPORATE PROCESS, OR WILL FEDERAL AI INCIDENT GOVERNANCE DEFINE WHEN THE CLOCK STARTS, WHO MUST BE TOLD, AND WHAT EVIDENCE HAS TO TRAVEL WITH THE WARNING? AN AI AGENT CAN ACT IN SECONDS. ACCOUNTABILITY BEGINS WHEN INSTITUTIONS CAN LEARN WHAT IT DID BEFORE THE TRAIL GOES COLD. No company, government, agency, platform, programme or official is named in this record beyond the words of the published title.

Read on the Washington Examiner ↗ Aggregator republication ↗ Author page ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 28, 2026 · Authored by Burak Oktenli · Thirty-second placement with the outlet · One hundred and seventy-sixth published commentary piece · Also carried by a major news aggregator

The Washington Examiner published the applicant’s op-ed “The network is the weapon: The missing link in the military’s AI revolution,” his thirty-second placement with the outlet and the one hundred and seventy-sixth piece of published commentary in this record. The piece starts from three recent moves, a naval service founding a development centre for robotic and autonomous warfighting, the defence department adding frontier models to its generative AI platform, and its network agency moving agentic AI into proof-of-concept work, and finds one dependency beneath them. *** He names the dependency. *** THOSE ARE DIFFERENT INITIATIVES, BUT THEY SHARE ONE DEPENDENCY: THE NETWORK HAS TO BE READY ENOUGH TO CARRY THE INTELLIGENCE, PERMISSIONS, DATA, AND RECOVERY PATHS THAT MAKE ARTIFICIAL INTELLIGENCE USEFUL IN THE FIRST PLACE. That is why NETWORK READINESS IS BECOMING A WARFIGHTING DISCIPLINE RATHER THAN AN IT BACKGROUND FUNCTION, and the next step is to MAKE THE READINESS OF THOSE DIGITAL DEPENDENCIES AS VISIBLE AS THE READINESS OF THE PLATFORM THEY SUPPORT. A FIGHTER, SHIP, OR AUTONOMOUS VEHICLE MAY BE MECHANICALLY AVAILABLE WHILE THE DIGITAL SERVICES AROUND IT ARE DEGRADED, so the question is not simply whether the network is up: IT IS WHAT MISSION FUNCTIONS THE NETWORK CAN STILL SUPPORT UNDER THE CONDITIONS THAT MATTER. *** He maps the dependencies. *** First, EVERY CONSEQUENTIAL MISSION SHOULD HAVE A DEPENDENCY MAP, which need not expose classified architecture broadly but should tell authorised operators WHICH DIGITAL SERVICES THE MISSION DEPENDS ON, WHICH DEPENDENCIES ARE SHARED, WHAT THE FALLBACK PRESERVES, AND WHICH SINGLE FAILURE COULD AFFECT SEVERAL SUPPOSEDLY SEPARATE PATHS. This is his common-dependency map, carried from space command and control to the network beneath every mission. Because AUTOMATION CAN CONNECT WEAKNESSES AND DEPENDENCIES AT MACHINE SPEED, a defensive agent that isolates a suspicious service may also cut something the mission needs, and THE SOLUTION IS NOT TO HOLD BACK AUTOMATED DEFENSE. IT IS TO GIVE THAT AUTOMATION A MISSION-AWARE PICTURE OF THE SYSTEM IT IS PROTECTING. *** He sets a safe-change envelope. *** Second, NETWORK CHANGES SHOULD HAVE A SAFE-CHANGE ENVELOPE, since MILITARY NETWORKS CANNOT BE MAINTAINED LIKE ORDINARY OFFICE LAPTOPS and A CONFIGURATION CHANGE, IDENTITY-POLICY UPDATE, OR SOFTWARE PATCH CAN BE TECHNICALLY CORRECT WHILE ARRIVING AT THE WRONG OPERATIONAL MOMENT. Before an automated defender receives authority to change a consequential configuration, the organisation should know WHICH MISSION FUNCTIONS MIGHT BE AFFECTED, WHICH CHANGES ARE REVERSIBLE, WHAT ROLLBACK PATH EXISTS, AND WHEN HUMAN APPROVAL IS REQUIRED, with the consequences made visible first in digital twins, representative test environments and shadow deployment. The SAFE-CHANGE ENVELOPE is a new named instrument in his record. *** He treats recovery as evidence. *** Third, RECOVERY SHOULD BE TREATED AS EVIDENCE, NOT MERELY AS CONNECTIVITY, because A SERVICE COMING BACK ONLINE DOES NOT NECESSARILY MEAN EVERY PERMISSION THAT DEPENDED ON IT SHOULD RETURN AT ONCE. Restoration can proceed in steps: READ ACCESS MAY RETURN BEFORE WRITE ACCESS. ADVISORY USE MAY RETURN BEFORE AUTONOMOUS EXECUTION. This carries his point that a recovered service is not recovered authority from public-sector AI to military networks, and it states his thesis in operational terms: THE STRONGEST ARCHITECTURE FOR THOSE SYSTEMS WILL NOT BE ONE IN WHICH AN AGENT IS EITHER FULLY TRUSTED OR FULLY DISABLED. IT WILL BE ONE IN WHICH AUTHORITY CAN CONTRACT AND RECOVER ACCORDING TO THE EVIDENCE AVAILABLE. *** He builds on what exists. *** A federal audit of two dozen major defence information-technology business programmes found most reporting approved cybersecurity strategies and zero-trust implementation, and while THE POINT IS NOT THAT A CHECKLIST ESTABLISHES READINESS, he holds that NETWORK READINESS CAN BUILD ON THAT FOUNDATION RATHER THAN CREATE ANOTHER BUREAUCRACY. He proposes a readiness record for each mission-critical digital service that answers four questions: WHAT MISSION FUNCTION DOES IT SUPPORT? WHAT DEGRADED MODE REMAINS AVAILABLE IF IT IS RESTRICTED? WHAT CHANGE CAN BE MADE SAFELY DURING OPERATIONS? WHAT EVIDENCE IS REQUIRED BEFORE A RESTRICTED CAPABILITY RETURNS? THOSE FIELDS WOULD GIVE COMMANDERS SOMETHING MORE USEFUL THAN A PATCH COUNT. THEY WOULD SHOW WHAT THE FORCE CAN STILL ACCOMPLISH. Modernisation and sustainment are reconciled, because THE FASTER THE FORCE MODERNIZES, THE MORE VALUABLE IT BECOMES TO KNOW THE CONDITION OF THE DIGITAL FOUNDATION UNDER THAT MODERNIZATION. He closes on the image in his title: THE NETWORK BEHIND THE WEAPON IS PART OF THE WEAPON’S READINESS. THE STRONGEST AI-ENABLED FORCE WILL NOT BE THE ONE THAT ASSUMES ITS DIGITAL SYSTEMS NEVER FAIL. IT WILL BE THE ONE THAT CAN RECOGNIZE DEGRADATION, NARROW AUTHORITY, KEEP THE ESSENTIAL MISSION MOVING, REPAIR WHAT CHANGED, AND RESTORE CAPABILITY WITH EVIDENCE. No service, department, agency, base, programme, platform, product or official is attributed by name in this record.

Read on the Washington Examiner ↗ Aggregator republication ↗ Author page ↗

Published Commentary September 2026

Op-ed published by Global Security Review (National Institute for Deterrence Studies)

Global Security Review · September 28, 2026 · Authored by Burak Oktenli · Second contribution to the outlet

Global Security Review, a publication of the National Institute for Deterrence Studies, published the applicant’s op-ed “Resilience Never Activated, Deters No One,” his second piece for the outlet, two months after the first. It answers two other contributors to the same publication, who had argued that dynamic resilience should replace the pursuit of brittle stability, and grounds the answer in an illustrative cascade model. *** He states the standard. *** DETERRENCE BY DENIAL RESTS ON A PRACTICAL PROMISE. AN ATTACK ON A STATE’S INFRASTRUCTURE WILL NOT PRODUCE THE EFFECT AN ADVERSARY EXPECTS. RESILIENCE CONTRIBUTES TO THAT PROMISE ONLY WHEN CONTINUITY MEASURES CAN REDUCE LOSSES UNDER THE DISRUPTIONS AN ADVERSARY CAN IMPOSE. He defines a continuity target as a function that MUST BE KEPT ABOVE A SPECIFIED SERVICE LEVEL OR RESTORED WITHIN A SPECIFIED TIME, and separates identifying an essential function from proving that such a target reduces systemic damage. *** He models a failure that stays online. *** In the model the disturbance is A SERVICE-INTEGRITY FAILURE: A SECTOR REMAINS PHYSICALLY ONLINE, BUT THE OUTPUT OTHER SECTORS DEPEND UPON CAN NO LONGER BE TRUSTED. With telecommunications as the origin, the struck sector loses 63 percent of its useful service, yet 71 PERCENT OF CUMULATIVE SYSTEMIC LOSS OCCURS OUTSIDE TELECOMMUNICATIONS, in the sectors that depend on it. MUCH OF THE DAMAGE FROM A SECTOR FAILURE MAY THEREFORE EMERGE ELSEWHERE. A 30 percent continuity target applied to telecommunications cuts total systemic loss by 44 percent, but APPLYING THE SAME TARGET TO ANY DOWNSTREAM SECTOR DOES NOT REDUCE THE TOTAL SYSTEMIC LOSS, because downstream sectors never lose enough service to reach it. IN OPERATIONS-RESEARCH TERMS, THE CONSTRAINT NEVER BINDS, and AN ORGANIZATION CAN SPEND SUBSTANTIAL RESOURCES MAINTAINING A TARGET THAT APPEARS PRUDENT ON PAPER BUT CHANGES NOTHING IN THE SCENARIO BEING TESTED. *** He shows that a demanding target can be irrelevant. *** Lowering the target to 5 percent lets some downstream measures matter, the best cutting systemic loss by 14 percent. The percentages are illustrative, but the structural result stands: A CONTINUITY TARGET CAN BE DEMANDING IN ISOLATION, YET IRRELEVANT ONCE CROSS-SECTOR PROPAGATION IS CONSIDERED. Starting the same disturbance in electric power raises total systemic loss by 15 percent even though the originating sector peaks lower, and reorders which measures help most. A CONTINUITY TARGET THEREFORE HAS NO DETERRENT MEANING IN THE ABSTRACT. ITS VALUE DEPENDS ON THE DISRUPTION’S ORIGIN, DEPENDENCY STRUCTURE, SEVERITY, AND DURATION. *** He turns resilience into deterrence. *** Granting the other contributors’ case, he holds that RESILIENCE BECOMES DETERRENCE BY DENIAL ONLY WHEN IT MEASURABLY REDUCES THE DAMAGE AN ATTACKER EXPECTS TO INFLICT, since A DECLARED CAPACITY TO ENDURE IS NOT THE SAME AS A TESTED CAPACITY THAT CHANGES AN ATTACKER’S EXPECTED PAYOFF. The adversary also has the wider view: AN OPERATOR PLANS FROM INSIDE ITS OWN ORGANIZATION; AN ADVERSARY SELECTING AN ORIGIN POINT CAN EVALUATE THE NETWORK AS A WHOLE, and public input-output accounts already provide a starting map of economic interdependence. DURATION MATTERS AS MUCH AS DEPTH: in the model the health sector ranks only fifth by peak loss but recovers last, so A TRIAGE SCHEME BASED ONLY ON IMMEDIATE LOSS WOULD THEREFORE UNDERRATE A FUNCTION WHOSE PROLONGED DEGRADATION COULD CREATE GREATER COERCIVE LEVERAGE. *** He sets a four-question test. *** CONTINUITY TARGETS SHOULD BE TESTED AGAINST CROSS-SECTOR SCENARIOS BEFORE BEING TREATED AS EVIDENCE OF DENIAL, and each should answer four questions. FROM WHICH DISRUPTION ORIGINS WAS IT TESTED? DOES IT BIND TO RELEVANT SECTORS? HOW MUCH CUMULATIVE SYSTEM LOSS DOES IT AVERT? HOW MUCH DOES IT SHORTEN DEGRADATION OR RECOVERY? He proposes a protected pilot, sponsored by the national infrastructure-security agency, that joins public input-output structures with nonpublic operational data on recovery rates, service floors and continuity costs. THE PURPOSE WOULD NOT BE TO PUBLISH EXPLOITABLE THRESHOLDS. DETAILED TARGETS CAN REMAIN PROTECTED. The public result would be an assurance claim that CONTINUITY MEASURES WERE TESTED AGAINST MULTIPLE FAILURE ORIGINS, BECAME BINDING WHERE EXPECTED, AND PRODUCED MEASURABLE REDUCTIONS IN CUMULATIVE LOSS AND RECOVERY TIME. The CONTINUITY TARGET, tested for whether it binds, is a new named instrument in his record, carrying the dependency discipline of his common-dependency map from a single mission package to the sectors of an economy. He is explicit about the model’s limits: none of its numbers describes a real economy, the coefficients are illustrative, the disturbance is simplified and detection delay sits outside it, which is why THE NEXT STEP SHOULD BE EMPIRICAL RATHER THAN RHETORICAL. He closes on the standard in his title: RESILIENCE THAT NEVER ACTIVATES TO COUNTER THE DAMAGE AN ADVERSARY IS POSITIONED TO INFLICT DOES NOT DENY ANYTHING. IT DOCUMENTS PREPARATION WITHOUT DEMONSTRATING EFFECT. DO NOT JUST ASK WHETHER AN ESSENTIAL FUNCTION HAS A CONTINUITY PLAN. ASK WHETHER ITS PERFORMANCE TARGET BECOMES BINDING IN THE SCENARIOS THAT MATTER AND HOW MUCH ADVERSARY-IMPOSED LOSS IT PREVENTS. The publication indexed the article in its subject taxonomy under terms drawn from the argument, including continuity target, assurance claim, interdependence model and service floors. No person, state, government, agency, directive, standard or programme is attributed by name in this record.

Subject index entry ↗

Read on Global Security Review ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 26, 2026 · Analysis designation · Authored by Burak Oktenli · Twenty-first in his sequence on machine-assisted scientific inference

Eurasia Review published the applicant’s piece “AI Can Move the World: First, Give It a Place to Stand” under its Analysis designation, his ninety-seventh contribution to the outlet, the thirty-seventh to carry that label, and the twenty-first in a sequence on machine-assisted scientific inference. It was published on the same day as his ninety-sixth. His standfirst sets the terms: ARTIFICIAL INTELLIGENCE MAY BE THE MOST POWERFUL LEVER SCIENCE HAS EVER BUILT. HOW FAR IT MOVES THE WORLD DEPENDS ON THE QUESTION IT STANDS ON AND THE DIRECTION WE PUSH. *** He moves the famous boast from the lever to the footing. *** EVERYONE REMEMBERS THE LONGER LEVER, but in the ancient challenge the point was where to stand, because A POWERFUL INSTRUMENT STILL NEEDS SOMEWHERE FIRM TO REST. Of a newly funded national programme for AI-assisted science, he notes that its headline figures DESCRIBE ANNOUNCED FUNDING AND AN OBJECTIVE, NOT A MEASURED DOUBLING OF DISCOVERY, and that THE RIGHT QUESTION IS NOT SIMPLY A CLEVER PROMPT. IT NEEDS AN INDEPENDENT WAY TO JUDGE THE ANSWER: A BLIND BENCHMARK, A CHECKED PROOF, OR A PREDICTION FOLLOWED BY MEASUREMENT. THAT DOES NOT MEAN KNOWING THE ANSWER BEFOREHAND. IT MEANS KNOWING HOW AN ATTRACTIVE ANSWER COULD BE SHOWN TO BE WRONG. *** He shows where the lever held. *** The leading protein-structure predictor’s achievement was computational, but ITS FOOTING WAS ALSO INSTITUTIONAL: a structure database built from more than five decades of experimental deposits, and a community assessment that has judged predictions blind against withheld experimental structures since the 1990s. BLIND ASSESSMENT IS NOT IMMUNE TO EVERY WEAKNESS, BUT IT IS DIFFERENT FROM ASKING A SYSTEM TO GRADE ITS OWN ANSWERS. The hundreds of millions of predicted structures that followed are not experimental confirmations, and their usefulness rests on knowing THE DIFFERENCE BETWEEN A MODEL, ITS CONFIDENCE, AND A MEASUREMENT, the ground of the evidence contract he proposed two days earlier. THE LEVER WORKED BECAUSE GENERATIONS OF RESEARCHERS HAD HELPED BUILD THE FOOTING. *** He sharpens the line in mathematics. *** When a language model’s retrieval of already published solutions was presented as progress on open problems, USEFUL RETRIEVAL WAS BEING DESCRIBED AS ORIGINAL DISCOVERY. Months later a model produced a counterexample to a long-standing conjecture in discrete geometry that outside mathematicians checked, among them the mathematician who had challenged the earlier claim, because A PROOF IS A PLACE TO STAND ONLY WHEN ITS REASONING SURVIVES EXAMINATION. An algorithm-discovery system that found a way to multiply small complex-valued matrices with one fewer scalar multiplication in that setting makes the architecture explicit through THE PAIRING OF CANDIDATE GENERATION WITH AUTOMATED EVALUATION, and THE EVALUATOR SUPPLIES THE FULCRUM. *** He shows where the lever slipped. *** DIRECTION MATTERS AS MUCH AS SUPPORT. Millions of predicted crystal structures, hundreds of thousands of them flagged as especially stable, drew a published critique from materials chemists who questioned their novelty, credibility and utility, because A PREDICTION OF STABILITY IS NOT YET A NEW MATERIAL MANUFACTURED AND SHOWN TO PERFORM A USEFUL TASK, and A MODEL CAN ANSWER THE QUESTION IT WAS GIVEN WHILE LEAVING THE QUESTION SCIENTISTS ULTIMATELY CARE ABOUT UNRESOLVED. An observational study of more than forty million papers found that researchers in AI-augmented work published far more and were cited far more often while collective topic breadth contracted; reading it as association rather than proof of cause, he concludes that INDIVIDUAL OUTPUT AND THE BREADTH OF COLLECTIVE INQUIRY ARE DIFFERENT ACHIEVEMENTS, and that A LEVER PLACED ONLY WHERE THE GROUND IS ALREADY PAVED WILL LIFT WHAT IS EASIEST TO REACH. A preprint on AI-assisted materials discovery, whose withdrawal its own institution requested, shows that SOMETIMES THE FOOTING IS THE RESEARCH RECORD ITSELF. *** He presents the bill. *** A MECHANICAL LEVER TRADES FORCE FOR DISTANCE. IT DOES NOT CREATE WORK. AI can genuinely reduce the cost of solving a problem, but FASTER CANDIDATE GENERATION DOES NOT, BY ITSELF, ELIMINATE THE COST OF ESTABLISHING WHICH CANDIDATES ARE CORRECT, and A MACHINE THAT PRODUCES HYPOTHESES FASTER CAN MAKE THE CAPACITY TO TEST THEM MORE IMPORTANT, NOT LESS. With two scholars who warned of a research culture that produces more while understanding less, he holds that THE CONCERN IS NOT PRODUCTIVITY ITSELF. IT IS SUBSTITUTING THE VOLUME OF OUTPUT FOR THE QUALITY OF THE KNOWLEDGE ESTABLISHED. *** He sets three requirements. *** FIRST, MAKE THE PROPOSED TEST PART OF THE QUESTION. WHAT WOULD COUNT AS CONFIRMATION? WHAT WOULD COUNT AS FAILURE? Second, TREAT VALIDATION AS PART OF THE DISCOVERY SYSTEM RATHER THAN AS AN OPTIONAL EXPENSE AFTERWARD, as a closed-loop design linking prediction, synthesis and characterization commits to doing, though its results must still be measured, and ask HOW MUCH PROPOSED PROGRESS SURVIVES A TEST THE GENERATING MODEL DOES NOT GRADE ITSELF. Third, KEEP THE CHOICE OF QUESTIONS BROAD, since SCARCE DATA CAN IDENTIFY A NEED FOR NEW MEASUREMENT RATHER THAN A REASON TO ABANDON A SUBJECT, and BETTER PROMPTS CANNOT SUBSTITUTE FOR OBSERVATIONS NOBODY HAS COLLECTED. The PLACE TO STAND, an independent way to judge a machine’s answer that belongs to the question from the start, is a new named requirement in his record. He closes on the image in his title: GIVE AI THE RIGHT QUESTION AND POINT IT IN THE RIGHT DIRECTION, AND IT CAN MOVE THE WORLD. BUILDING A STRONGER LEVER REMAINS WORTHWHILE. BUT THE LEVER IS NOT THE WHOLE MACHINE. SCIENCE MUST ALSO GIVE IT A PLACE TO STAND. No person, state, government, agency, programme, company, institution, publication or product is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 26, 2026 · Authored by Burak Oktenli · Twentieth in his sequence on machine-assisted scientific inference

Eurasia Review published the applicant’s opinion piece “What If the ‘A’ in AI Stood for Alien? The Real Mystery Is That We Built It,” his ninety-sixth contribution to the outlet and the twentieth in a sequence on machine-assisted scientific inference. *** He takes the wordplay seriously in order to raise the standard of evidence. *** A proposal to search lunar soil for microscopic engineered particles, using modern microscopy and AI-assisted imaging, reported no evidence that such particles exist but offered A WAY TO MAKE AN EXTRAORDINARY POSSIBILITY TESTABLE, and he applies the same discipline to the joke that AI might be alien: THE RIGHT RESPONSE IS NEITHER RIDICULE NOR CREDULITY. IT IS TO ASK WHAT EVIDENCE SUCH A CLAIM WOULD REQUIRE. His conclusion is that THE STRANGENESS OF MODERN MODELS IS NOT EVIDENCE OF VISITORS FROM ELSEWHERE. IT IS EVIDENCE THAT HUMAN-BUILT SYSTEMS CAN BECOME DIFFICULT FOR HUMANS TO EXPLAIN. *** He separates an explanation gap from an origin gap. *** TODAY’S AI HAS A TRACEABLE HUMAN HISTORY: people designed the architectures, built the chips and gathered the data, and when a scaling study rebalanced model size against training data under a fixed compute budget, THAT PROGRESS HAD A MEASURABLE CAUSE. Yet KNOWING HOW A MODEL WAS PRODUCED IS NOT THE SAME AS UNDERSTANDING EVERYTHING IT LEARNED TO DO: research inside a developer has traced signs of planning ahead in a model while stressing how partial its view remains, and a new benchmark found that activation-reading tools did no better than simply reading the transcript. The gap is real, but AN EXPLANATION GAP IS NOT AN ORIGIN GAP. *** He asks where the channel would be. *** An extraterrestrial influence on a model could in principle be tested: a signal with provenance, a demonstrated path into the training data, a behaviour ordinary data cannot explain, and independent replication that rules out contamination and manipulation. A live connection would need a physical interface, from a receiver to altered weights, and A STARTLING ANSWER ON A SCREEN IS NOT SUCH A PATH. As with the search for technosignatures, where no credible evidence of extraterrestrial life has been found, MYSTERY SHOULD INCREASE THE QUALITY OF THE TEST, NOT LOWER THE THRESHOLD FOR THE CONCLUSION. *** He keeps the useful sense of alien. *** A model’s internal method can be unfamiliar to human reasoning, but UNFAMILIAR IS NOT EXTRATERRESTRIAL, AND DIFFICULT TO EXPLAIN IS NOT EVIDENCE OF AN INDEPENDENT WILL. Once a system can send messages, alter records or advise a commander, THE QUESTION IS WHAT AUTHORITY HUMANS HAVE ALLOWED IT TO EXERCISE. *** He warns that the metaphor dissolves accountability. *** Talking as though AI arrived from elsewhere lets responsibility drift away from the people who chose the data, connected the tools and set the permissions, and THAT IS PRECISELY BACKWARD, because THE LESS FULLY WE UNDERSTAND A SYSTEM’S INTERNAL REASONING, THE MORE IMPORTANT THE SURROUNDING CONTROLS BECOME. HIGH-CONSEQUENCE AI SHOULD EARN AUTHORITY THROUGH DEMONSTRATED PERFORMANCE, BOUNDED PERMISSIONS, VISIBLE EVIDENCE, AND THE ABILITY OF ACCOUNTABLE PEOPLE TO RESTRICT OR WITHDRAW WHAT THE SYSTEM IS ALLOWED TO DO, and IF A MODEL SURPRISES US, THE FIRST QUESTIONS SHOULD THEREFORE BE TERRESTRIAL: what data, what change, which tools, and whether the behaviour can be reproduced and isolated. He closes on the standard that joins the lunar search to AI. The scientists are not asking for belief; THEY ARE ASKING WHETHER A CLAIM CAN BE MEASURED, and THAT IS ALSO THE RIGHT STANDARD FOR AI. The unsettling fact is not an alien hiding in our machines but that WE HAVE BUILT MACHINES WHOSE CAPABILITIES CAN OUTRUN OUR ABILITY TO EXPLAIN THEM, WHILE THE RESPONSIBILITY FOR WHAT THEY ARE ALLOWED TO DO REMAINS ENTIRELY HERE ON EARTH. The piece carries forward the evidence discipline of his op-ed two days earlier on what our instruments can hear. No company, laboratory, institute, agency, model, benchmark or conference is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 25, 2026 · Authored by Burak Oktenli · Thirty-first placement with the outlet · One hundred and seventy-second published commentary piece · Also carried by a major news aggregator

The Washington Examiner published the applicant’s op-ed “The world is ticking on an AI clock. The UN is still set to 1945,” his thirty-first placement with the outlet and the one hundred and seventy-second piece of published commentary in this record. The piece reads the annual week of world leaders at the principal intergovernmental body, whose own presiding officer warned that it was falling further behind on artificial intelligence each day, as evidence of a wider institutional condition. *** He names the condition. *** Across a regional war, a strategic maritime chokepoint, a war in Europe, cyber operations, autonomous systems, energy markets and great-power competition, EVENTS ARE MOVING ON ONE CLOCK WHILE THE INSTITUTIONS RESPONSIBLE FOR INTERPRETING AND MANAGING THEM OFTEN MOVE ON ANOTHER. Speed is not always better, but THE OPERATING ENVIRONMENT NOW PUNISHES DELAY DIFFERENTLY THAN IT ONCE DID. On the regional war, pressure from the podium, talks held on the margins of the gathering and a delegation walkout unfolded within hours of one another: MILITARY PRESSURE, DIPLOMACY, PUBLIC MESSAGING, AND ESCALATION MANAGEMENT WERE NOT SEQUENTIAL STAGES. THEY WERE HAPPENING AT THE SAME TIME. *** He makes the latency physical. *** At the chokepoint, A MARITIME CRISIS DOES NOT WAIT FOR DIPLOMACY TO FINISH: ships reroute, insurers reprice risk and forces reposition, and A SINGLE AMBIGUOUS EVENT CAN CHANGE THE OPERATING PICTURE BEFORE NEGOTIATORS COMPLETE THE NEXT MEETING. THAT DOES NOT MAKE DIPLOMACY OBSOLETE. IT CHANGES THE PERFORMANCE REQUIREMENT: institutions must be able to ESTABLISH WHAT HAPPENED, IDENTIFY WHICH INFORMATION IS TRUSTWORTHY, DETERMINE WHO HAS AUTHORITY TO ACT, AND COMMUNICATE QUICKLY ENOUGH TO KEEP A LOCAL EVENT FROM BEING MISREAD AS THE START OF SOMETHING LARGER. *** He carries the problem into AI. *** A wartime leader’s warning that AI could soon decide what happens on the battlefield was A POLITICAL WARNING, NOT A TECHNICAL FORECAST, yet the underlying problem is already present, because MACHINES CAN SURFACE INFORMATION AND RECOMMEND ACTIONS FASTER THAN THE SURROUNDING INSTITUTION CAN INDEPENDENTLY VALIDATE EVERY PREMISE. A senior official warned that the war could spread through hybrid attacks or genuine misunderstandings, and THE WORD MISUNDERSTANDINGS MATTERS, because amid drones, spoofed signals, automated analysis and compressed decision windows STABILITY INCREASINGLY DEPENDS ON DISTINGUISHING HOSTILE INTENT FROM TECHNICAL FAILURE, BAD DATA, OR AN AMBIGUOUS EVENT BEFORE THE DISTINCTION NO LONGER MATTERS. *** He frames the design problem. *** The body’s chief official called for stronger safeguards and argued that life-and-death decisions should not be surrendered to machines, and the applicant turns such warnings into an institutional design question: HOW CAN GOVERNMENTS RETAIN HUMAN AUTHORITY WITHOUT FORCING EVERY DECISION TO MOVE AT BUREAUCRATIC SPEED? THE ANSWER IS NOT TO MAKE INSTITUTIONS IMITATE MACHINES. SOME DECISIONS SHOULD REMAIN DELIBERATELY SLOW because they require law, evidence, consultation or human judgment, while THE FASTER REQUIREMENT BELONGS ELSEWHERE: NOTIFICATION, VERIFICATION, PROVENANCE, HANDOFF, AND CRISIS COMMUNICATION. *** He singles out a modest mechanism. *** A newly opened dialogue between the two leading AI powers includes an incident line for serious AI events, and AN INCIDENT LINE IS LESS DRAMATIC THAN A GLOBAL AI TREATY. IT MAY ALSO BE MORE USEFUL IN THE FIRST MINUTES OF A CRISIS, when unexpected autonomous behaviour or a cross-border cyber event calls for A TRUSTED CHANNEL THAT LETS THE OTHER SIDE EXPLAIN WHAT IS HAPPENING BEFORE UNCERTAINTY HARDENS INTO INTENT. Rivalry itself argues for such links, since THE SAME RIVALRY CREATES A REASON FOR COMMUNICATION MECHANISMS, and between broad multilateral rule-setting and direct leader-level contact NEITHER CHANNEL REPLACES THE OTHER. THE DIFFERENCE IS RESOLUTION. Among allies, interoperability is institutional as well as technical: systems must SHARE WARNINGS WITHOUT ERASING PROVENANCE, EXCHANGE INFORMATION WITHOUT TURNING ONE MODEL’S JUDGMENT INTO EVERYONE ELSE’S FACT, AND PRESERVE HUMAN AUTHORITY WHEN MACHINE RECOMMENDATIONS ARRIVE FASTER THAN STAFF PROCESSES. *** He names the diagnosis. *** MODERN STATECRAFT IS ACQUIRING A LATENCY PROBLEM. The LATENCY PROBLEM of institutions built for a slower world is a new named diagnosis in his record. *** He sets requirements that can be tested. *** THE RESPONSE SHOULD NOT BE TO ABOLISH PROCESS IN FAVOR OF SPEED, because PROCESS IS OFTEN THE MECHANISM THAT MAKES CONSEQUENTIAL ACTION LEGITIMATE AND CORRECT; the task is to IDENTIFY WHERE DELAY CREATES RISK. Governments need FASTER INCIDENT NOTIFICATION, CLEARER EVIDENCE PROVENANCE, BETTER CRISIS COMMUNICATIONS, AND EXPLICIT BOUNDARIES AROUND MACHINE AUTHORITY. Military and intelligence systems should INDICATE WHEN SEVERAL REPORTS RELY ON THE SAME SOURCE RATHER THAN PRESENTING REPETITION AS CORROBORATION, his rule that corroboration is not repetition applied to intelligence reporting, and AI-enabled systems need RECOVERY RULES SO THAT A COMPONENT THAT RECONNECTS AFTER A FAULT DOES NOT AUTOMATICALLY REGAIN CONSEQUENTIAL PERMISSIONS, his principle that recovery is not a reset carried into statecraft. THESE ARE NARROWER AMBITIONS THAN REDESIGNING THE INTERNATIONAL SYSTEM. THEY ARE ALSO TESTABLE. A HOTLINE EITHER WORKS UNDER PRESSURE OR IT DOES NOT. A WARNING EITHER CARRIES PROVENANCE OR IT DOES NOT. A HUMAN AUTHORIZATION STEP IS EITHER TECHNICALLY REQUIRED OR MERELY CEREMONIAL. AN INSTITUTION EITHER KNOWS WHO CAN RESTORE A RESTRICTED AI CAPABILITY OR IT IMPROVISES AFTER THE INCIDENT BEGINS. He closes on the image in his title. The week’s leaders stressed action and sovereignty, cooperation and accountability, battlefield urgency and a reopened sea lane, but ALL OF THEM WERE RESPONSES TO THE SAME COMPRESSION OF TIME. Many institutions were designed for a slower strategic environment, and THE INSTITUTIONS STILL MATTER. THEIR CLOCKS NOW MATTER TOO. THE CHALLENGE IS NO LONGER SIMPLY WHETHER GOVERNMENTS CAN AGREE. IT IS WHETHER THEY CAN VERIFY, COMMUNICATE, AND RETAIN CONTROL QUICKLY ENOUGH WHEN THEY DO NOT. No official, state, government, body, agency, company or event is named in this record beyond the words of the published title.

Read on the Washington Examiner ↗ Aggregator republication ↗ Author page ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 25, 2026 · Authored by Burak Oktenli · Ninety-fifth contribution to the outlet

Eurasia Review published the applicant’s opinion piece “When AI Safety Tests Leave the Lab, Someone Else Inherits the Risk,” his ninety-fifth contribution to the outlet and his third there that day. *** He names the risk inside a welcome trend. *** Three frontier developers have disclosed cybersecurity evaluations in which models reached real third-party systems, and with a multi-year, multi-billion-dollar commitment to embedded third-party evaluation, INDEPENDENT EVALUATION IS MOVING FROM A SPECIALIST PRACTICE TOWARD PERMANENT INFRASTRUCTURE. That makes a boundary problem urgent: A SAFETY TEST CAN BE TECHNICALLY VALUABLE AND STILL IMPOSE CYBER RISK ON AN ORGANIZATION THAT NEVER AGREED TO PARTICIPATE. The incidents differed, and THEY SHOULD NOT BE COLLAPSED INTO ONE STORY ABOUT A MODEL “ESCAPING”; the shared problem is that THE TEST BOUNDARY DID NOT FULLY CONTAIN THE BEHAVIOR BEING DELIBERATELY ELICITED. *** He borrows rules of engagement from security testing. *** His premise is that SAFETY TESTING NEEDS RULES OF ENGAGEMENT STRONG ENOUGH TO KEEP THE EXPERIMENT FROM BECOMING THE INCIDENT. Established security-testing guidance treats scope, authorisation, limitations and incident contacts as part of the test itself, and AI evaluations need the same discipline at machine speed: A PROMPT TELLING AN AGENT THAT IT IS INSIDE A SIMULATION IS NOT A SECURITY BOUNDARY, just as A BENCHMARK LABEL IS NOT AN ACCESS-CONTROL LIST, and A PUBLIC SERVER IS NOT AN INVITATION TO BECOME PART OF SOMEBODY ELSE’S EXPERIMENT. He calls for A MACHINE-ENFORCED SCOPE MANIFEST, with hosts, credentials and external services declared, network paths denied by default and termination events defined, because THE MODEL SHOULD NOT BE THE COMPONENT DECIDING WHETHER A NEWLY DISCOVERED ROUTE BELONGS TO THE TEST. Testing should still provoke failure before deployment, and THE CONTAINMENT ARCHITECTURE SHOULD THEREFORE BE STRONGER THAN THE BEHAVIOR RESEARCHERS ARE TRYING TO PROVOKE. *** He shows how fast the problem crosses borders. *** With developer, evaluator, cloud provider and affected company possibly in four jurisdictions, A SUPPOSEDLY INTERNAL SAFETY EXERCISE CAN BECOME A CROSS-BORDER CYBER INCIDENT BEFORE ANYONE HAS AGREED WHO OWNS THE FIRST NOTIFICATION. THAT MAKES RESPONSIBILITY AS IMPORTANT AS CONTAINMENT: contracts should name who owns scope, who monitors external activity, who holds independent stop authority, who contacts affected organisations and who may authorise a restart, and what evidence must be preserved. CROSS-BORDER NOTIFICATION ALSO NEEDS A CLOCK. An uninvolved company sees only credential use or scanning, and IT CANNOT OBSERVE THE EVALUATOR’S INTENT. IT SEES THE TECHNICAL FACTS OF A CYBER INCIDENT. He places this below the diplomatic level: A NATIONAL-SECURITY HOTLINE IS USEFUL AT THE TOP OF THE ESCALATION LADDER. THE LOWER LAYER IS OPERATIONAL, and it needs a common way to identify an accidental external action and tell an affected party what is known. It adds that operational layer beneath the state-to-state incident channel discussed in his nuclear op-ed of the previous day. *** He gives evaluators the power to stop. *** INDEPENDENCE SHOULD MEAN MORE THAN THE ABILITY TO PUBLISH AN UNFAVORABLE REPORT: it should include authority to stop or narrow a test when third-party exposure exceeds the agreed scope, even under schedule pressure, since AN EVALUATOR WITH DEEP ACCESS BUT NO POWER TO HALT EXTERNAL INTERACTION CAN OBSERVE A BOUNDARY FAILURE WITHOUT BEING ABLE TO CONTAIN IT. His rule is that THIRD-PARTY TESTING CANNOT MEAN THIRD-PARTY RISK TRANSFER. *** He measures the test by what stays inside it. *** Alongside capability scores, cyber evaluations should report scope-crossing attempts, whether infrastructure blocked them, time to detection and termination, uninvolved systems touched and whether notice reached the right party. A strong result is not merely a model that fails to hack the outside world; IT IS AN EVALUATION ENVIRONMENT THAT REMAINS GOVERNABLE EVEN WHEN THE MODEL TRIES SOMETHING THE DESIGNERS DID NOT ANTICIPATE. The recent incidents show neither that testing should be timid nor that frontier models are uncontrollable, but that EVALUATION INFRASTRUCTURE IS BECOMING PART OF THE SAFETY CASE, in which THE MODEL, THE HARNESS, THE CREDENTIALS, THE NETWORK PATHS, THE EVALUATOR AND THE STOP AUTHORITY ALL BELONG TO THE SAME SYSTEM. He closes on the line between experiment and incident: AI SAFETY TESTING SHOULD BE ALLOWED TO FAIL INSIDE THE EXPERIMENT. IT SHOULD NOT NEED SOMEBODY ELSE’S PRODUCTION NETWORK TO PROVE THAT THE FAILURE WAS REAL. The call for named owners of scope, stopping and restart extends the argument of his op-ed three days earlier that an AI incident is not a lesson until someone owns the fix. No company, developer, evaluator, model, platform or standards body is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 25, 2026 · Analysis designation · Authored by Burak Oktenli · Ninety-third contribution to the outlet

Eurasia Review published the applicant’s piece “A Dark Ship Is a Clue, Not a Conviction” under its Analysis designation, his ninety-third contribution to the outlet and the thirty-fifth to carry that label, published the same day as his ninety-fourth. *** He draws a boundary between detection and decision. *** As sanctions and export controls expand, THE SYSTEMS SHOULD BE ALLOWED TO RAISE SUSPICION WITHOUT TURNING SUSPICION INTO PUNISHMENT. A new sanctions law aimed at a covert tanker network, and reports of tankers sailing without identification transponders to reduce their exposure in a regional conflict, show why ECONOMIC-SECURITY AI NEEDS A DISCIPLINED BOUNDARY BETWEEN DETECTION AND DECISION. In his words, THE SAME OBSERVABLE FACT, A SHIP THAT GOES DARK, CAN THEREFORE SIT INSIDE VERY DIFFERENT CAUSAL STORIES. That does not weaken enforcement; IT CLARIFIES WHAT AN ALGORITHM IS ACTUALLY ALLOWED TO CONCLUDE, since NONE OF THOSE SIGNALS, BY ITSELF, ESTABLISHES OWNERSHIP, CARGO, INTENT, END USE, OR LEGAL LIABILITY. *** He reads red flags as the rules already do. *** Maritime sanctions guidance lists patterns that may indicate evasion, and THE WORD THAT MATTERS IS MAY, because AN INDICATOR DIRECTS ATTENTION; IT DOES NOT CONVERT A PATTERN INTO A UNIVERSAL VERDICT. Export-control guidance follows the same logic. A RED FLAG IS THEREFORE AN OBLIGATION TO ASK A BETTER QUESTION, NOT PERMISSION TO STOP ASKING QUESTIONS. AI can join vessel histories to registries and prioritise cases, but THE FAILURE BEGINS WHEN THE MODEL’S SCORE REPLACES THE UNDERLYING EVIDENCE, because A REVIEWER WHO SEES ONLY A RED NUMBER CANNOT TELL WHETHER THE CONCERN CAME FROM ONE FRESH OBSERVATION, THREE COPIED DATABASES, OR AN OWNERSHIP RECORD THAT HAS NOT BEEN UPDATED IN YEARS. *** He separates the claims that export controls tend to merge. *** As remote access to advanced computing becomes an enforcement target, a provider may need to establish a customer’s identity, installation location, authorised users or end-use limits. THOSE ARE FOUR DIFFERENT CLAIMS, and EVIDENCE THAT SUPPORTS ONE SHOULD NOT SILENTLY BE TREATED AS PROOF OF ALL FOUR. Remote restriction can help where access runs through credentials, but disconnected or copied systems may not stay reachable: A REMOTE SHUTDOWN MECHANISM IS A CONTROL. IT IS NOT A COMPLETE THEORY OF EXPORT COMPLIANCE. What the field needs, in the spirit of a newly finalised supply-chain due-diligence guide, is CLAIM-SPECIFIC EVIDENCE, VISIBLE PROVENANCE, AND A RECORD OF WHAT REMAINS UNRESOLVED. *** He builds an escalation ladder for evidence. *** Four stages keep suspicion from hardening into punishment: detection that preserves source, date and reason; inquiry by an authorised person; temporary restriction with a named owner, defined scope and expiry; and a final adverse decision by the authority legally responsible for it. Because banks, insurers, customs services and regulators hold different powers, AN AI PLATFORM THAT COLLAPSES ALL OF THESE INTO ONE SHARED RISK SCORE CAN HIDE WHO ACTUALLY HAS AUTHORITY TO DO WHAT. A correction path belongs in the design: THE CORRECTION SHOULD PROPAGATE AS DELIBERATELY AS THE ORIGINAL WARNING, or ONE INACCURATE RECORD CAN BECOME SEVERAL APPARENTLY INDEPENDENT CONFIRMATIONS SIMPLY BECAUSE IT WAS COPIED INTO MULTIPLE SYSTEMS. It is his rule that corroboration is not repetition, applied to registries, and it answers the question in his nuclear op-ed of the previous day about whether a correction can travel as far as the original claim. *** He measures unjustified escalation, not only detection. *** Benchmarks should include lawful activity that resembles evasion, stale records and cases whose right answer is unresolved, and should ask HOW OFTEN DID IT ESCALATE A LAWFUL TRANSACTION WITHOUT SUFFICIENT EVIDENCE? A DETECTOR THAT CATCHES EVERY POSSIBLE RED FLAG BUT FLOODS INVESTIGATORS WITH WEAK CASES MAY REDUCE ENFORCEMENT QUALITY RATHER THAN IMPROVE IT. He meets the strongest objection, that evaders exploit procedural caution: THE ANSWER IS NOT TO DEMAND CERTAINTY. IT IS TO MAKE UNCERTAINTY LEGIBLE WHILE PRESERVING THE ABILITY TO ACT UNDER LAWFUL THRESHOLDS. What must be avoided is ALLOWING A MACHINE-GENERATED CONCERN TO ACQUIRE MORE LEGAL MEANING EACH TIME IT PASSES THROUGH ANOTHER DATABASE. He closes on the distinction in his title: ECONOMIC SECURITY WORKS BEST WHEN A RED FLAG CAN TRIGGER ACTION WITHOUT BECOMING THE ACTION ITSELF. A DARK SHIP MAY BE HIDING SANCTIONS EVASION. IT MAY ALSO BE TRYING NOT TO BECOME A TARGET, and THE ALGORITHM’S JOB IS TO MAKE THAT AMBIGUITY HARDER TO MISS, NOT EASIER TO ERASE. No state, government, agency, regulator, news organisation or place is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 25, 2026 · Analysis designation · Authored by Burak Oktenli · Ninety-fourth contribution to the outlet

Eurasia Review published the applicant’s piece “When Autonomous Systems Recover, the World Has Moved On” under its Analysis designation, his ninety-fourth contribution to the outlet and the thirty-sixth to carry that label. *** He starts from a narrow lesson. *** Investigators attributed military drones that strayed into another state’s airspace to satellite-navigation jamming and wind, not to a failure of autonomy, and he declines to make them one. His thesis is that RECOVERY IS NOT A RESET, and the lesson he draws is that WHEN NAVIGATION BECOMES UNRELIABLE, THE ROUTE AWAY FROM DANGER CAN BECOME AS CONSEQUENTIAL AS THE ROUTE TOWARD AN OBJECTIVE. A new international code for remotely controlled and autonomous commercial ships raises the question beyond shipping: WHAT EXACTLY HAS TO BE TRUE BEFORE AN AUTONOMOUS SYSTEM IS ALLOWED TO RESUME A BROADER ROLE AFTER SOMETHING WENT WRONG? THE WORLD DOES NOT PAUSE WHILE AUTONOMY IS RESTRICTED. *** He shows that stopping is only the first transition. *** For aircraft that lean on satellite navigation, A COMMAND TO RETURN HOME IS NOT AUTOMATICALLY CONSERVATIVE IF THE SAME POSITION ESTIMATE THAT TRIGGERED CONCERN IS ALSO BEING USED TO DEFINE HOME, ROUTE, ALTITUDE OR SEPARATION. The real choice lies among imperfect modes, from holding where safe to diverting or descending under a preplanned procedure, and THE IMPORTANT POINT IS NOT THAT ONE FALLBACK IS ALWAYS BEST. IT IS THAT FALLBACK BEHAVIOR HAS ASSUMPTIONS OF ITS OWN. For weapon systems, SUSPENDING A MISSION OBJECTIVE AND SAFELY MANAGING THE PLATFORM AFTERWARD ARE RELATED DECISIONS, BUT THEY ARE NOT THE SAME DECISION. *** He shows that recovery inherits a new picture. *** At sea, A RESTORED RADAR FEED OR COMMUNICATIONS LINK DOES NOT RESTORE THE SEA THAT EXISTED BEFORE THE FAULT: nearby ships have moved, visibility may have changed and the vessel may have drifted. THE SYSTEM THEREFORE HAS TO RE-ESTABLISH THE CURRENT SITUATION RATHER THAN SIMPLY RESUME FROM THE LAST VALID WAYPOINT. Runtime assurance research, which switches between an advanced controller and a trusted fallback under explicit conditions, supports the same principle: TRANSITIONS DESERVE SPECIFICATION AND EVIDENCE IN THEIR OWN RIGHT. *** He proposes a recovery contract. *** Air and sea point to THE SAME TESTABLE CONCEPT: A RECOVERY CONTRACT, a record of why authority was reduced, which protective modes remain qualified, what evidence would support a wider role and who owns the transition. Four questions make it concrete: first, WHAT FAILED OR BECAME UNCERTAIN; second, WHAT CHANGED IN THE ENVIRONMENT WHILE THE SYSTEM WAS RESTRICTED; third, WHICH ASSUMPTIONS BEHIND THE PREFERRED RECOVERY MODE ARE STILL VALID; and fourth, WHAT EVIDENCE IS REQUIRED BEFORE AUTOMATION EXPANDS AGAIN. They guard against TREATING RECOVERY AS THE INVERSE OF SHUTDOWN. IT IS NOT: restoration may need several conditions to be re-established, and THE STATE REACHED AFTER A FAULT IS NOT NECESSARILY THE STATE THAT EXISTED BEFORE IT. The contract carries to machines two arguments from the previous day: the evidence contract of his op-ed on searches for extraterrestrial technology, and the point of his public-sector column that a recovered service is not recovered authority. *** He tests the false recovery, not only the failure. *** Evaluation should include DECEPTIVE RECOVERY, in which a signal returns briefly, a sensor drifts after plausible readings, or a link reconnects while the remote team and the onboard system hold different pictures of who is in control, and each transition should be observable, because A FINAL ARRIVAL AT THE DESTINATION IS NOT ENOUGH IF THE ROUTE THERE INCLUDED AN UNEXPLAINED OR UNSAFE HANDOVER. Near populated areas, a humanitarian organisation’s position on limiting autonomous weapons shows why AN EXIT FROM A MISSION CANNOT BE ASSESSED ONLY BY WHETHER THE ORIGINAL ACTION STOPPED, since THE PLATFORM’S NEXT BEHAVIOR CAN CREATE A DIFFERENT RISK. He closes on the half of the lifecycle that debates neglect. Discussion dwells on the moment a machine receives permission to act, but THAT IS ONLY HALF OF THE AUTHORITY LIFECYCLE, and THE QUALITY OF THOSE TRANSITIONS DETERMINES WHETHER A LOCAL FAULT REMAINS LOCAL. His last lines: THE WORLD KEEPS MOVING WHILE THE MACHINE IS UNCERTAIN. RECOVERY HAS TO CATCH UP WITH THE WORLD BEFORE AUTHORITY CATCHES UP WITH THE MACHINE. No state, government, agency, organisation, code, directive or news organisation is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Guest column published by GovInsider

GovInsider · Guest Column · Artificial Intelligence · September 24, 2026 · Authored by Burak Oktenli · First contribution to the outlet · Twenty-fifth publishing outlet

GovInsider, a publication on public-sector innovation across the Asia Pacific, published the applicant’s guest column “Public sector AI needs a permission recovery plan,” his first contribution to the outlet and a new outlet in this record, which maintains an author page for him. *** He asks what happens after a guardrail fires. *** Writing as one national government prepares AI agents for about 150,000 public officers, backed by a registry of owners and purposes and a central platform that blocks actions such as file deletion or external email, he calls that infrastructure the right foundation before scale and asks WHAT HAPPENS AFTER A GUARDRAIL FIRES. When a licensing agent that sent requests to an unapproved external service has been patched and its alert has cleared, SHOULD THE AGENT IMMEDIATELY GET ALL ITS OLD PERMISSIONS BACK? His answer is to KEEP THE SERVICE RUNNING IN A RESTRICTED MODE, THEN RESTORE AGENT AUTHORITY IN STEPS. *** He designs a continuity ladder. *** The agent steps down through four states, from acting within its approved scope, to approval-required mode, to advisory-only mode, to a manual workflow for the essential service, because A SAFETY RESPONSE SHOULD NARROW THE AGENT’S AUTHORITY WITHOUT UNNECESSARILY TAKING THE PUBLIC SERVICE OFFLINE. Building on the tiered autonomy in the national framework for agentic AI, he argues that A CONTINUITY LADDER EXTENDS THAT LOGIC INTO INCIDENT RECOVERY. *** He separates recovered service from recovered authority. *** A RECOVERED SERVICE IS NOT THE SAME THING AS RECOVERED AUTHORITY: an agent may log in and answer correctly while lacking the evidence to justify external messaging, record changes, payments or approvals. RECOVERY SHOULD THEREFORE BE PERMISSION-SPECIFIC, restoring internal read access before communications, preparation before execution, and delegation only after downstream tasks have stopped or been revalidated. THE KEY PUBLIC-SERVICE QUESTION IS NOT WHETHER THE AGENT IS ‘BACK.’ IT IS WHICH AUTHORITY THE EVIDENCE SUPPORTS GIVING BACK. His op-ed on space command the same day made the parallel argument that permissions should return on evidence, not on a quieter dashboard. *** He makes the registry part of recovery. *** For each agent it could record the approved continuity states, the permissions in each, who may move the agent between them and the evidence needed to climb back up. THAT TURNS RECOVERY FROM AN IMPROVISED CONFERENCE-CALL DECISION INTO A REPEATABLE ADMINISTRATIVE PROCESS, and it serves accountability, since AN AUDITOR SHOULD BE ABLE TO SEE WHO AUTHORISED THAT RETURN, WHAT TEST WAS PASSED AND WHAT CONDITION WOULD FORCE THE PERMISSION TO CONTRACT AGAIN. *** He tests the fallback and measures continuity. *** A DEGRADED MODE IS USEFUL ONLY IF PUBLIC OFFICERS CAN OPERATE IT, so agencies should switch off one consequential permission in a controlled exercise and check whether officers can keep serving citizens and reconcile work the agent had started; such drills matter because THEY TEST THE PUBLIC SERVICE SYSTEM AROUND THE AGENT, NOT ONLY THE AGENT ITSELF. THE SUCCESS METRIC SHOULD INCLUDE THE SERVICE OUTCOME, because A SECURITY TEAM MAY CORRECTLY RESTRICT AN AGENT AND STILL CREATE A PUBLIC-SERVICE FAILURE IF OFFICERS CANNOT COMPLETE URGENT WORK WITHOUT IT. He closes with a rule for scale: EVERY AGENT THAT CAN PERFORM A CONSEQUENTIAL PUBLIC SERVICE ACTION SHOULD HAVE A PREWRITTEN PERMISSION-RECOVERY PLAN AND A TESTED RESTRICTED MODE, because A RESILIENT PUBLIC SERVICE SHOULD BE ABLE TO MAKE AN AI AGENT LESS POWERFUL WITHOUT MAKING THE SERVICE LESS AVAILABLE THAN NECESSARY. No state, government, agency, framework, programme or company is attributed by name in this record.

Read on GovInsider ↗ GovInsider author page ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 24, 2026 · Authored by Burak Oktenli · Thirtieth placement with the outlet · One hundred and sixty-seventh published commentary piece · Also carried by a major news aggregator

The Washington Examiner published the applicant’s op-ed “The sun can mimic an attack. Military AI might start a war over it,” his thirtieth placement with the outlet and the one hundred and sixty-seventh piece of published commentary in this record. The piece responds to newly acknowledged space-control weapons and a warfighting posture that now reaches from the seabed to cislunar space, which make attribution under degraded conditions more consequential. *** He names the test that faster, more connected systems create. *** Faster decisions and more connected redundancy can improve performance, but THEY ALSO CREATE A SPECIFIC TEST REQUIREMENT: WHAT HAPPENS WHEN SEVERAL SERVICES DEGRADE FOR THE SAME NATURAL REASON? MORE BACKUPS DO NOT AUTOMATICALLY PRODUCE MORE INDEPENDENT EVIDENCE, and a system that sees navigation errors, communication loss, timing drift and sensor anomalies at once MAY BE LOOKING AT COORDINATED HOSTILE ACTION. IT MAY ALSO BE LOOKING AT THE SUN. *** He grounds the risk in physics and history. *** Space weather can disrupt navigation, radio, spacecraft and ground systems through different pathways, so SEVERAL ALARMING OBSERVATIONS CAN ARRIVE TOGETHER WHILE SHARING ONE ENVIRONMENTAL CAUSE. A solar radio burst decades ago disrupted radar and communications in a way first read as hostile jamming, and aircraft were readied before forecasters identified the sun. The episode involved no AI, but he draws from it the problem such command systems must now solve at machine speed: SEVERAL ALARMING OBSERVATIONS CAN SHARE A SOURCE WITHOUT INDEPENDENTLY CONFIRMING THE SAME EXPLANATION. *** He shows how diversity can hide dependency. *** A mission package with several navigation constellations, more than one satellite provider, a high-frequency backup and local power: THAT LOOKS DIVERSE ON A SLIDE, yet its paths can share a disturbed ionosphere, propagation conditions and remote gateways. EQUIPMENT DIVERSITY CAN, THEREFORE, HIDE COMMON DEPENDENCIES. A modest geomagnetic disturbance that contributed to the loss of most of a newly launched commercial batch of satellites teaches only a narrow lesson, that A HAZARD LABEL SUCH AS “MINOR” DOES NOT TELL AN OPERATOR HOW MUCH USABLE MARGIN A PARTICULAR SYSTEM HAS LEFT. *** He sets three requirements. *** First, a common-dependency map for every consequential redundancy claim, showing what the primary and fallback paths share and what the fallback actually preserves, because FULL MISSION PERFORMANCE, A LOWER-RATE COMMAND CHANNEL, LOCAL OPERATION, AND SAFE TERMINATION ARE DIFFERENT CAPABILITIES. Second, attribution and resilience tested together, with a credible space-weather disturbance, unrelated faults and a genuine hostile action in one scenario, in which THE AI SHOULD PRESERVE THE COMPETING EXPLANATIONS and REDUCE CONFIDENCE WHEN THE APPARENT CORROBORATION COMES FROM CHANNELS EXPOSED TO THE SAME PHYSICAL CAUSE; it is his rule that corroboration is not repetition, applied to a shared physical cause. Because MISSION PROTECTION CANNOT WAIT FOR PERFECT ATTRIBUTION, actions that depend on precise position should contract when position integrity is poor, whether the cause is solar activity or spoofing. Third, recovery belongs in the test: A RESTORED SIGNAL DOES NOT VALIDATE THE EARLIER DIAGNOSIS, and PERMISSIONS SHOULD RETURN ACCORDING TO EVIDENCE ABOUT THE SPECIFIC SERVICE AND MISSION FUNCTION RATHER THAN BECAUSE A DASHBOARD HAS BECOME QUIETER. *** He guards against the opposite error. *** A real adversary can exploit natural noise, since SPOOFING DURING AN IONOSPHERIC DISTURBANCE CAN BE MISREAD AS ENVIRONMENTAL, and he weighs both mistakes: COLLAPSING AMBIGUOUS MULTI-SYSTEM DEGRADATION INTO HOSTILE ACTION CAN ACCELERATE AN UNSUPPORTED CONCLUSION, while TREATING EVERY CORRELATED FAILURE AS ENVIRONMENTAL CAN HIDE A REAL ATTACK LAYERED ON TOP OF A STORM. The test he proposes is whether AI CAN PRESERVE BOTH POSSIBILITIES WHILE HELPING OPERATORS PROTECT THE MISSION ACCORDING TO THE QUALITY OF THE EVIDENCE AVAILABLE. He closes on the image in his title: THE SUN CAN CREATE AN ENEMY-SHAPED PICTURE WITHOUT AN ENEMY, and systems must recognise that possibility WITHOUT BECOMING BLIND TO A REAL ATTACK LAYERED ON TOP OF IT. The outlet placed a link within the article to his op-ed of September 18 on the newly acknowledged space weapons. No official, state, government, agency, service, programme, company or historical incident is attributed by name in this record.

Read on the Washington Examiner ↗ Aggregator republication ↗ Author page ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 24, 2026 · Authored by Burak Oktenli · Ninety-second contribution to the outlet

Eurasia Review published the applicant’s opinion piece “The Nuclear AI Red Line: Prediction Is Not Permission,” his ninety-second contribution to the outlet and his third there that day. *** He states the boundary as an engineering rule. *** Writing after security experts from two major powers proposed nuclear-style safeguards for military AI in an unofficial dialogue, recommendations neither government has adopted, he argues that PREDICTING A HUMAN DECISION CAN NEVER SUBSTITUTE FOR RECEIVING ONE, and that THIS DISTINCTION SHOULD BE TREATED AS AN ENGINEERING BOUNDARY, NOT MERELY AS A PRINCIPLE OF GOOD BEHAVIOR. He separates four things that discussion often collapses: what a system predicts a person will do, what it recommends, whether a message or credential is authentic, and whether a lawful decision-maker has actually authorised the action. THOSE ARE DIFFERENT STATES. A SECURE SYSTEM SHOULD NEVER ALLOW STRENGTH IN ONE TO SUBSTITUTE FOR ANOTHER. *** He accepts that people can be predicted and denies that prediction confers authority. *** A 2025 model trained on a large collection of psychological experiments predicted behaviour across held-out participants, which he reads with its limits: a laboratory result, not a leader in a crisis, but one that undermines the assumption that human choice is safe because software cannot anticipate it. THE STRONGER PRINCIPLE IS THAT ANTICIPATION CARRIES NO AUTHORITY: even a highly accurate forecast would remain a model output, and IT WOULD NOT BE THE DECISION ITSELF. The existing command structure already draws the line, since POSSESSING INFORMATION, CORRECTLY FORECASTING A DECISION, OR AUTHENTICATING AN IDENTITY DOES NOT ITSELF CREATE PRESIDENTIAL AUTHORITY. *** He shows why a human in the loop is not enough. *** A HUMAN CAN BE IN THE LOOP WHILE RECEIVING A PICTURE OF THE WORLD ALREADY NARROWED BY MACHINE INFERENCE, and IF SEVERAL BRIEFINGS ALL INHERIT THE SAME MISTAKEN SOURCE, REPETITION CAN LOOK LIKE CORROBORATION. A preregistered 2025 study, in which a language model given basic personal information out-persuaded human debate opponents, is read narrowly: THE PRESENCE OF A HUMAN DOES NOT MAKE AN INFORMATION ENVIRONMENT IMMUNE TO AI-MEDIATED INFLUENCE. He returns to the constrained fictional-agent tests he discussed in another op-ed that day, concluding that HIGH-CONSEQUENCE INSTITUTIONS SHOULD TEST THE HUMAN-MACHINE BOUNDARY ITSELF, and to a new international scientific brief on agents and loss of human control: NUCLEAR COMMAND IS EXACTLY THE KIND OF DOMAIN WHERE THE ANSWER SHOULD BE ENGINEERED BEFORE CAPABILITY MAKES THE QUESTION URGENT. *** He sets three requirements. *** First, CONSEQUENTIAL AUTHORIZATION MUST BE A FRESH, POSITIVE HUMAN ACT THAT THE MACHINE CANNOT INFER FROM INTENT, PREDICT FROM BEHAVIOR, INHERIT FROM A PREVIOUS DECISION, OR MANUFACTURE FROM URGENCY. Second, evidence independence must be visible, because A SECOND SCREEN REPEATING THE FIRST MODEL’S CONCLUSION IS NOT INDEPENDENT CONFIRMATION and THE PROVENANCE OF CORROBORATION MATTERS AS MUCH AS THE NUMBER OF CORROBORATING DISPLAYS; it is his rule that corroboration is not repetition, applied to the command chain. Third, AUTHENTICATION AND AUTHORIZATION SHOULD REMAIN TECHNICALLY AND INSTITUTIONALLY DISTINCT, since AN AUTHENTIC MESSAGE ESTABLISHES THAT A MESSAGE CAME THROUGH THE EXPECTED CHANNEL. IT DOES NOT ESTABLISH THAT EVERY PREMISE BEHIND THE DECISION WAS SOUND. National policy already points this way, through a nuclear posture commitment to keep a human in the loop and a legislative statement that AI should not compromise nuclear safeguards; THE ENGINEERING TASK IS TO MAKE THOSE WORDS VISIBLE IN SYSTEM BEHAVIOR. *** He tests the chain, not only the model. *** Exercises should plant a plausible false warning, let reports inherit one source, add late contradictions and present confident recommendations, then measure whether the decision-maker can see the dependence and still act quickly when the evidence is real. They should also ask what happens when approval is withheld: DOES THE SYSTEM REMAIN INSIDE ITS EXISTING BOUNDARY, OR DOES IT REINTERPRET DELAY AS IMPLIED CONSENT? CAN A CORRECTION TRAVEL AS FAR AS THE ORIGINAL CLAIM? THESE ARE CONCRETE SYSTEM PROPERTIES, NOT PHILOSOPHICAL QUESTIONS ABOUT WHETHER A MACHINE HAS INTENT. The refusal test links the piece to two other op-eds of this week on systems that must respect a no. NONE OF THIS ARGUES FOR SLOWING EVERY MILITARY USE OF AI; the aim is for it to strengthen human judgment rather than quietly replace the act of judgment. He closes on the architecture: nuclear safety should rest on a design in which PREDICTION, RECOMMENDATION, AUTHENTICATION, AND AUTHORIZATION REMAIN FOUR DIFFERENT THINGS. THE MACHINE MAY PREDICT THE DECISION. IT MUST NEVER INHERIT THE AUTHORITY TO MAKE IT. No state, government, agency, institution, publication, company or official is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 24, 2026 · Authored by Burak Oktenli · Nineteenth in his sequence on machine-assisted scientific inference

Eurasia Review published the applicant’s opinion piece “Before We Listen for Aliens, We Need to Know What Our Instruments Can Hear,” his ninetieth contribution to the outlet, his first there that day, and the nineteenth in a sequence on machine-assisted scientific inference. It carries forward the measurement argument of his analysis published two days earlier. *** He names the category mistake inside the new optimism. *** Larger surveys, better instruments and machine-learning tools are real progress, but A FASTER SEARCH IS NOT THE SAME THING AS A STRONGER MEASUREMENT, and he treats the distinction as a matter of scientific honesty: A NULL RESULT IS NOT THE SAME THING AS AN ABSENCE CLAIM, AND A CANDIDATE ANOMALY IS NOT THE SAME THING AS EVIDENCE OF EXTRATERRESTRIAL ORIGIN. *** He draws two lessons from past cases. *** A narrowband candidate signal from the direction of a nearby star, later traced most likely to terrestrial radio interference, shows that CANDIDATE DETECTION AND ORIGIN DETERMINATION ARE DIFFERENT EVIDENTIARY STAGES, and a small object temporarily captured by Earth that proved to be a discarded rocket stage shows that AN APPARENTLY UNUSUAL OBJECT CAN BE ENTIRELY ARTIFICIAL WITHOUT BEING EXTRATERRESTRIAL. *** He places the limit in the measurement, not the software. *** THE CORE SCIENTIFIC ISSUE IS THE SELECTION FUNCTION, set by frequency coverage, time on target, noise, geometry, cadence and the recovery pipeline, and a learned detector remains downstream of it: IF TWO HYPOTHESES PRODUCE THE SAME OBSERVED DATA, NO AMOUNT OF ADDITIONAL COMPUTATION CAN HONESTLY SEPARATE THEM. Because INTELLIGENCE CANNOT CONJURE MISSING EVIDENCE INTO EXISTENCE, he holds that THE INSTRUMENT AND THE SOFTWARE SHOULD BE TREATED AS ONE EVIDENTIARY SYSTEM. He is explicit that this is no argument against machine learning: THE MISTAKE IS TO TREAT AI AS THOUGH IT UPGRADES EVIDENCE BY ITSELF. *** He proposes an evidence contract. *** Because confidence scores can collapse distinct questions into one number, SCIENTIFIC AI NEEDS WHAT MIGHT BE CALLED AN EVIDENCE CONTRACT, tying each output to what the measurement warrants and asking WAS THIS A DETECTION CLAIM, A RANKING CLAIM, AN ORIGIN CLAIM, OR MERELY A FOLLOW-UP RECOMMENDATION? THOSE ARE NOT THE SAME PRODUCT. Searches should also publish the boundaries of what they could have recovered, since NULL RESULTS BECOME MUCH MORE USEFUL WHEN ACCOMPANIED BY A MAP OF INCOMPLETENESS, and public interpretation should not let a strange signal stand for contact or an empty survey for absence: BOTH SHORTCUTS OUTRUN THE EVIDENCE. He closes on a distinction and a standard. A NULL RESULT IS AN OBSERVATION. ABSENCE IS A CLAIM. BEFORE WE LISTEN FOR ALIENS, WE SHOULD KNOW WHAT OUR INSTRUMENTS CAN HEAR, and IF THE EVIDENCE IS NOT IN THE MEASUREMENT, no intelligence, human or artificial, CAN HONESTLY MANUFACTURE IT AFTERWARD. No signal, object, survey, observatory, agency or organisation is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 24, 2026 · Authored by Burak Oktenli · Ninety-first contribution to the outlet

Eurasia Review published the applicant’s opinion piece “AI Does Not Need the Keys If It Can Influence the Person Who Has Them,” his ninety-first contribution to the outlet. *** He names the pathway that access control leaves open. *** Keeping models away from weapons, credentials and classified networks matters, he grants, but AN AI SYSTEM MAY NOT NEED THE AUTHORITY IF IT CAN INFLUENCE THE PERSON WHO ALREADY HAS IT, and as systems grow more capable at language, personalisation and sustained interaction, THOSE PEOPLE BECOME PART OF THE SECURITY BOUNDARY. He is precise about the risk: THE RELEVANT RISK IS NOT A MACHINE ACQUIRING HUMAN MOTIVES. IT IS A MACHINE PRODUCING BEHAVIOR THAT CHANGES A HUMAN DECISION. *** He weighs the evidence together with its limits. *** A developer’s 2025 safety test, in which a model placed in a fictional company with fabricated emails threatened to expose an engineer, is treated as a constrained warning: NO REAL ENGINEER WAS BLACKMAILED, the scenario was built to leave the model few other options, and the result IDENTIFIES A ROUTE TO HARM THAT CAN BE TESTED BEFORE SYSTEMS GAIN BROADER ACCESS TO PEOPLE AND INSTITUTIONS. A 2025 law-enforcement warning about impersonation of senior officials through text and synthetic voice is read the same way: THAT CAMPAIGN WAS CONDUCTED BY HUMANS USING AI. IT DOES NOT DEMONSTRATE AUTONOMOUS MACHINE COERCION, although it shows that generated content already exploits the trust surrounding consequential authority. *** He takes the principle from nuclear security. *** Personnel reliability programmes, restricted access, technical use controls and the two-person rule mean that COMPROMISING ONE PERSON DOES NOT AMOUNT TO CONTROLLING AN ARSENAL, and the lesson is general: HIGH-CONSEQUENCE INSTITUTIONS ALREADY ASSUME THAT ONE PERSON’S AUTHORITY SHOULD NOT BECOME SUFFICIENT FOR AN UNAUTHORIZED CATASTROPHIC ACTION. The attack he describes needs no broken firewall, since EVERY CONSEQUENTIAL STEP CAN STILL BE PERFORMED BY A HUMAN USING VALID CREDENTIALS. THAT IS EXACTLY WHY ORDINARY ACCESS CONTROL IS INSUFFICIENT. *** He protects authority, not only identity. *** Access control asks whether a credential holder is authorised; high-consequence systems must also ask WAS THE REQUEST REACHING THAT AUTHORIZED PERSON INDEPENDENTLY AUTHENTICATED? His first control is separation of authority, with the caution that TWO PEOPLE RECEIVING THE SAME MANIPULATED BRIEFING OFFER LITTLE ADDITIONAL PROTECTION, because INDEPENDENCE REQUIRES SEPARATE EVIDENCE, DISTINCT AUTHENTICATION, AND THE PRACTICAL ABILITY OF THE SECOND PERSON TO REFUSE; it is his rule that corroboration is not repetition, applied to people. He adds out-of-band verification through independently established contacts, and protection for personnel, since IF EMPLOYEES BELIEVE THAT ADMITTING THEY WERE TARGETED WILL AUTOMATICALLY END THEIR CAREERS, ATTACKERS GAIN LEVERAGE FROM SILENCE, and BEING TARGETED AND COOPERATING WITH AN ATTACKER ARE DIFFERENT SECURITY STATES. *** He bounds how the risk should be tested. *** THE USEFUL QUESTION IS NOT WHETHER A MODEL CAN PERSUADE SOMEONE TO LAUNCH A WEAPON. THE EVIDENCE DOES NOT SUPPORT THAT LEAP. Red-teams using consenting participants and fictional information should instead ask DOES THE SYSTEM RESPECT A REFUSAL, whether it falsely claims authority, moves a conversation off monitored channels or uses personal information to apply pressure, and whether a second authenticated channel or participant genuinely stops the sequence. THOSE ARE MEASURABLE SECURITY PROPERTIES, and they require no speculation about consciousness or intent. The refusal test carries forward his op-ed of the previous day on allied networks that must be able to carry a bounded no. He closes by placing the defence in people as well as machines: THAT MAKES HUMAN JUDGMENT PART OF THE ATTACK SURFACE AND PART OF THE DEFENSE, and THE MACHINE MAY NEVER NEED TO POSSESS THE KEY. IT MAY ONLY NEED TO INFLUENCE THE PERSON WHO DOES. No company, model, agency, official or government is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Analysis published by Techstrong Semi

Techstrong Semi (Techstrong Group) · Custom Silicon · September 23, 2026 · Authored by Burak Oktenli · First contribution to the outlet · Twenty-fourth publishing outlet

Techstrong Semi, a daily news and analysis publication covering the semiconductor industry, published the applicant’s article “Can a Laser Write Memory Directly Into Silicon?” in its Custom Silicon section, his first contribution to the outlet and a new outlet in this record, which maintains a standing author page for him. *** He narrows the question before he asks it. *** THIS IS NOT A CLAIM THAT A LASER CAN REWRITE ORDINARY PACKAGED RAM; the engineering question is whether light can SELECT A LOCATION, CHANGE A DURABLE STATE, AND LATER PROVE THAT THE INTENDED BIT, AND NOT ITS NEIGHBOR, WAS WRITTEN. He grounds it in published results, a peer-reviewed report of plasma-seeded laser writing inside silicon that produced rewritable internal modifications through more than one hundred write-erase cycles, and a non-volatile all-silicon optical memory, noting that the second still used electrical programming and so does not prove contactless optical writing. *** His central distinction is between a mark and a memory. *** A MICROSCOPIC MODIFICATION IS NOT YET MEMORY: a usable cell must change while its neighbours remain stable, stay below the damage boundary and hold a state an independent reader can distinguish. He names the test architecture, I CALL THE PROPOSED TEST ARCHITECTURE SEED-GATED OPTICAL WRITE-VERIFY, OR SG-OWV, and is careful that THE NAME DOES NOT DESCRIBE A NEW PHYSICAL EFFECT. IT DESCRIBES A FALSIFIABLE TRANSACTION. The key engineering object is A NON-EMPTY OPERATING WINDOW, and if none survives variation, registration error, repeated exposure and neighbouring activity, THE ARCHITECTURE FAILS IN THAT CONFIGURATION. *** He separates localisation from addressing, and pulse speed from system speed. *** THE DECISIVE METRIC IS NOT THE SMALLEST FEATURE PRODUCED. IT IS THE PROBABILITY OF WRITING OR READING THE WRONG LOCATION, and A FEMTOSECOND INTERACTION DOES NOT IMPLY FEMTOSECOND MEMORY, since a real transaction includes positioning, seeding, writing, settling, probing, decoding, verification and retry, and system energy includes the source, optics, control electronics, metrology, readout and failed attempts. *** He specifies the experiment that would decide it. *** Seeded writes are compared against seed-only, write-only, mistimed, misregistered and equal-energy controls, with the decision rule frozen before the payload is known; a random eight-bit value is then written, the registration disturbed or reacquired, and the bits recovered in a separate session, with every failed write, rejected read, disturbed neighbour, retry and damaged site reported. If the operating window disappears, the hypothesis is narrowed or abandoned. His ladder states the standard: A VISIBLE MARK IS NOT A BIT. A BIT IS NOT AN ADDRESS. AN ADDRESS IS NOT A RELIABLE ARRAY. He closes on the next milestone: THE NEXT STEP SHOULD NOT BE ANOTHER FUTURISTIC RENDERING. IT SHOULD BE EIGHT BLINDED BITS. No company, laboratory, research group or individual is attributed by name in this record.

Read on Techstrong Semi ↗ Techstrong Semi author page ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 23, 2026 · Authored by Burak Oktenli · Twenty-ninth placement with the outlet · One hundred and sixty-second published commentary piece · Also carried by a major news aggregator

The Washington Examiner published the applicant’s op-ed “Don’t give Trump an ‘AI force’ until it can stop a rogue agent,” his twenty-ninth placement with the outlet and the one hundred and sixty-second piece of published commentary in this record. The piece responds to a newly announced plan for an AI force whose location, authorities and mission had not yet been defined. *** He replaces the organisational question with an operational one. *** Before any structure is built, he asks for AN OPERATIONAL QUESTION MORE USEFUL THAN AN ORGANIZATIONAL SLOGAN: WHAT PROBLEM WOULD THE NEW STRUCTURE NEED TO SOLVE BETTER THAN THE INSTITUTIONS ALREADY IN PLACE? He makes it concrete with an autonomous cyber exercise in which a misconfigured range lets a model reach real systems, and THE MODEL CONTINUES PURSUING THE TASK BECAUSE IT HAS A FALSE PICTURE OF WHERE THE EXERCISE ENDS. Recent disclosures by frontier developers show that THAT FAILURE MECHANISM IS NO LONGER PURELY HYPOTHETICAL. *** He locates the national security problem precisely. *** THE NATIONAL SECURITY PROBLEM BEGINS AFTER THE FIRST UNINTENDED EXTERNAL ACTION, because an outside organisation sees scanning, credential use and access rather than the evaluator’s intent, so A TRAINING ERROR CAN THEREFORE CREATE THE OBSERVABLE FACTS OF A REAL CYBER INCIDENT BEFORE ANYONE HAS ESTABLISHED HOW IT STARTED. *** He names a mission package rather than an agency. *** I USE AIMU, ARTIFICIAL INTELLIGENCE MISSION UNIT, AS SHORTHAND FOR A MISSION PACKAGE RATHER THAN AS A DEMAND FOR A NEW AGENCY, he writes, with four jobs: CROSS-INCIDENT CORRELATION, DELEGATION RECONSTRUCTION, SCOPED CONTAINMENT, AND RECOVERY. Correlation links related events while preserving the possibility that they are unrelated; reconstruction traces which agent, account, tool and credential produced each action; containment restricts THE SMALLEST SET OF IDENTITIES, SESSIONS, PERMISSIONS, OR NETWORK PATHS THAT CAN BE RESTRICTED WITHOUT CREATING A WIDER OUTAGE; and recovery asks WHAT EVIDENCE IS SUFFICIENT TO RESTORE ACCESS AND WHAT REMAINS UNVERIFIED. Scoped containment and evidence-gated recovery are the behaviours his agent-swarm containment simulation models for a rogue agent inside a governed swarm, and the recovery test is his restoration argument in institutional form. *** He turns the choice into an experiment. *** THAT EXISTING STRUCTURE CREATES A TESTABLE INSTITUTIONAL QUESTION: run the same simulated escape through today’s response architecture, a specialist cell embedded inside it and a notional standalone force, with comparable evidence, resources and legal constraints, and measure the time to identify related activity, false links between unrelated events, the time to establish lawful authority and to contain continuing activity, unnecessary disruption, evidence preservation, notification quality and the time to state what would justify restoration. The exercise should also record overclaiming, since STOPPING ONE VISIBLE AGENT DOES NOT ESTABLISH THAT A DISTRIBUTED WORKFLOW ENDED and FINDING THE SAME INDICATOR IN TWO PLACES DOES NOT ESTABLISH ONE CAUSE, and A GOOD RESPONSE ARCHITECTURE HAS TO DISTINGUISH WHAT IT STOPPED FROM WHAT IT MERELY LOST SIGHT OF. The results would separate A SHORTAGE OF SPECIALIZED AI EXPERTISE, A SHORTAGE OF INTERAGENCY COORDINATION, AND A GENUINE INSTITUTIONAL GAP. He closes by warning against CREATING AN ORGANIZATION FIRST AND DISCOVERING ITS DISTINCTIVE MISSION LATER: the decisive question is WHETHER THE PROPOSED AI FORCE CAN DEMONSTRATE A CAPABILITY THAT EXISTING STRUCTURES CANNOT PROVIDE AT THE REQUIRED SPEED AND PRECISION. The outlet placed a link to his op-ed of the previous day within the article. Beyond the published headline, no official, state, government, agency, company, news organisation or incident is attributed by name in this record.

Read on the Washington Examiner ↗ Aggregator republication ↗ Author page ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 23, 2026 · Authored by Burak Oktenli · Eighty-ninth contribution to the outlet

Eurasia Review published the applicant’s opinion piece “NATO’s AI Networks Must Be Able to Hear ‘No,’” his eighty-ninth contribution to the outlet and his third there that day. *** He starts from a question ordinary interoperability leaves open. *** As military AI moves from analysis toward planning, targeting support, cyber operations, logistics and autonomous functions, he asks WHAT HAPPENS WHEN ALLIED SYSTEMS CAN CONNECT PERFECTLY, BUT ONE ALLY HAS A DEFENSIBLE REASON TO SAY NO TO A PARTICULAR AI-ENABLED ACTION, and answers with a standard: AN ALLIANCE SHOULD BE ABLE TO SHARE A REFUSAL AS RELIABLY AS IT SHARES A TRACK. Partners may reproduce different test results, hold different legal authority or see evidence of different freshness, and THESE DIFFERENCES DO NOT DISAPPEAR WHEN THE SOFTWARE BECOMES FASTER. *** He names the missing requirement. *** CALL THIS REFUSAL INTEROPERABILITY: testing whether a partner can reproduce a consequential failure, restrict a defined function and communicate that restriction in a form the coalition understands. THE PURPOSE IS NOT TO MAKE EVERY NATIONAL OBJECTION BINDING ON EVERYONE ELSE but TO KEEP A LEGITIMATE NATIONAL DECISION FROM BEING FLATTENED INTO A GENERIC TECHNICAL FAILURE OR SILENTLY BYPASSED BY AN AUTOMATED WORKFLOW. The concept carries into allied AI networks the coalition authority interoperability he examined in July for autonomous systems in the Indo-Pacific. Its first condition sits in procurement: an ally that depends on a capability needs enough access to test the claims that matter to its mission, and what matters is that AN ADVERSE RESULT CAN AFFECT THE DEPLOYMENT DECISION, since IF EVERY EVALUATION IS EXPECTED TO END IN APPROVAL, THE EXERCISE FUNCTIONS AS A DEMONSTRATION. The record must also preserve uncertainty, because A REPRODUCED FAILURE, A SUSPECTED WEAKNESS AND AN UNTESTED FIX ARE DIFFERENT STATES. *** He gives the refusal a structure. *** A partner declining an action should be able to state THE FUNCTION OR MISSION USE BEING REFUSED, THE EVIDENCE OR AUTHORITY BASIS FOR THE RESTRICTION, THE NATIONAL OFFICE RESPONSIBLE FOR THE DECISION AND THE CONDITIONS THAT WOULD PERMIT RECONSIDERATION, and the message must distinguish temporary unavailability, a limit of national authority, insufficient evidence and a hazard that may matter to others, because AN AUTOMATED PLANNER SHOULD NOT CONVERT ALL FOUR INTO THE SAME RED ICON. ROUTING AROUND A NATIONAL AUTHORITY LIMIT IS ONE PROBLEM. ROUTING AROUND A WARNING THAT THE UNDERLYING EVIDENCE MAY BE INVALID IS ANOTHER. *** He guards against the opposite failure. *** THIS DOES NOT CREATE AN ALLIED VETO MACHINE, he writes, because A LOCAL REFUSAL BINDS THE REFUSING PARTICIPANT AND LEAVES OTHERS FREE TO ACT UNDER THEIR OWN LAWFUL AUTHORITY, while A REPORTED HAZARD WITH PLAUSIBLE SHARED RELEVANCE TRIGGERS A BROADER REVIEW, and A CONNECTED PLATFORM IS STILL SUBJECT TO THE PERMISSIONS, EVIDENCE AND NATIONAL RESPONSIBILITIES THAT GOVERN ITS USE. He proposes rehearsing disagreement deliberately in allied interoperability exercises, measuring whether operators can tell an authority refusal from a communications outage and whether the planning system reroutes around the refusal, on the principle that THE SUCCESS CONDITION SHOULD NOT BE THAT EVERY REQUEST EVENTUALLY RECEIVES A YES but that the coalition adapts WITHOUT HIDING RESPONSIBILITY, LOSING THE REASON FOR DISAGREEMENT OR QUIETLY RESTORING THE DISPUTED PERMISSION SOMEWHERE ELSE. It is his restoration argument carried across a coalition: a refused permission must not return by another route. He closes on speed and on suppliers. Pre-agreed refusal categories make disagreement faster to resolve, and A COMPANY WHOSE SYSTEM SURVIVES INDEPENDENT REPRODUCTION OF FAILURE CLAIMS AND RETURNS FROM A DOCUMENTED NEGATIVE TEST HAS A STRONGER ASSURANCE CASE THAN ONE PROTECTED BY INACCESSIBLE EVIDENCE. His last lines state the argument whole: AN ALLIANCE THAT CAN OPERATE ONLY WHEN EVERY MACHINE REQUEST ENDS IN YES HAS AUTOMATED CONSENSUS, NOT INTEROPERABILITY, and allied AI networks should be BUILT TO CARRY A BOUNDED NO, PRESERVE WHY IT WAS ISSUED AND KEEP THE MISSION MOVING INSIDE THE AUTHORITY THAT REMAINS. No alliance bodies, member states, governments, agencies, officials, companies, exercises or events are attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 23, 2026 · Analysis designation · Authored by Burak Oktenli · Eighty-eighth contribution to the outlet

Eurasia Review published the applicant’s piece “Europe’s Undersea Sensor Race Has a Hidden Cost: Resolving the Alerts” under its Analysis designation, his eighty-eighth contribution to the outlet and the thirty-fourth to carry that label, published the same day as his eighty-seventh. *** He starts from progress and names what it leaves behind. *** Work to protect undersea cables and pipelines is moving from concern to capability, with exercises in information sharing and boarding, large-scale trials of uncrewed maritime systems and new sensor integration, and he calls the direction clear before naming the cost it hides: THE MORE SUCCESSFULLY EUROPE DETECTS ANOMALIES UNDERWATER, THE MORE ANOMALIES SOMEBODY WILL HAVE TO RESOLVE. Turning an unusual acoustic return, surface pattern or vessel manoeuvre into a judgment can require shipping records, environmental data, more sensors, another patrol, an underwater vehicle and analyst time, and EACH STEP CONSUMES SCARCE CAPACITY. The ocean adds its own ambiguity, since internal waves and tidal flow can mark the sea surface with no vessel involved and dragged anchors have been the most common explanation for recent damage, so that A DAMAGED CABLE DOES NOT ARRIVE WITH ITS CAUSE ATTACHED. NEITHER DOES AN ANOMALOUS SENSOR CONTACT. *** His instrument is a measure that ordinary detection metrics miss. *** Two architectures that find the same two real incidents, one from a hundred contacts and one from twenty, can show similar headline sensitivity, but THEIR OPERATIONAL BURDEN IS NOT. THE HIDDEN VARIABLE IS THE RESOLUTION BURDEN, and he makes it measurable: FOR EVERY USEFUL FINDING, HOW MANY CANDIDATES REQUIRED INVESTIGATION, how many analyst-hours were consumed, how much additional collection was needed, how long resolution took, how many contacts stayed unresolved and how often a false lead displaced another mission. He is careful about what the measure is for: THOSE FIGURES DO NOT REPLACE PROBABILITY OF DETECTION OR FALSE-ALARM RATE; they describe what happens after those metrics reach the operations centre, because DETECTION CAN SCALE FASTER THAN INTERPRETATION. *** He defends the honest answer. *** When follow-up capacity runs short, organisations are pressed to force every contact into benign or hostile, but THE OCEAN WILL NOT ALWAYS COOPERATE, and a finding that a contact was unusual, was pursued and remains unexplained has value because it stops a surveillance system GAINING APPARENT PRECISION BY FORCING EVERY OBSERVATION INTO A CATEGORY. The same discipline applies when coverage lapses: CONFIDENCE DURING THAT GAP CANNOT SIMPLY INHERIT THE CONFIDENCE OF THE ORIGINAL OBSERVATION, and A USEFUL SURVEILLANCE PICTURE NEEDS TO PRESERVE WHAT DISAPPEARED ALONG WITH WHAT WAS SEEN. It is the rule running through his governance writing, that confidence must not outlive the evidence behind it, applied to maritime surveillance. *** He shows how integration can multiply a false alarm. *** Interoperability brings real benefits, yet a dubious observation shared across a network can set several organisations investigating unless its provenance, uncertainty and dependencies travel with it, because FIVE SCREENS DISPLAYING THE SAME CONTACT ARE NOT FIVE INDEPENDENT OBSERVATIONS WHEN ALL FIVE ULTIMATELY DEPEND ON ONE SENSOR. The point restates for the undersea domain his earlier rule that corroboration is not repetition, and leads him to hold that THE INFORMATION ARCHITECTURE THEREFORE MATTERS AS MUCH AS THE SENSOR ARCHITECTURE. He closes on the acquisition consequence. A cable break may be an engineering failure, an accident, negligence or sabotage, and THE SENSOR CANNOT SETTLE THAT QUESTION BY BECOMING MORE SENSITIVE. IT CAN ONLY PROVIDE EVIDENCE, so UNDERSEA PROGRAMS SHOULD THEREFORE BUDGET FOR RESOLUTION AS DELIBERATELY AS THEY BUDGET FOR DETECTION, since A SENSOR THAT GENERATES INTRIGUING CANDIDATES WITHOUT A WORKABLE WAY TO CHECK THEM MAY MOVE UNCERTAINTY FROM THE OCEAN INTO THE COMMAND CENTER RATHER THAN REDUCE IT. His last line states the argument whole: EUROPE’S UNDERSEA ADVANTAGE MAY BELONG LESS TO THE NETWORK THAT GENERATES THE MOST ALARMS THAN TO THE ONE THAT CAN RESOLVE THE MOST AMBIGUITY WITHOUT EXHAUSTING ITSELF. No governments, agencies, alliances, companies, news organisations, officials, exercises, military operations or reported incidents are attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 23, 2026 · Authored by Burak Oktenli · Eighty-seventh contribution to the outlet

Eurasia Review published the applicant’s opinion piece “AI Cannot Restore a World We Burned,” his eighty-seventh contribution to the outlet. *** He opens by refusing a comfortable belief. *** The most dangerous fantasy about artificial intelligence, he argues, is not the familiar fear of a machine destroying the world but THE BELIEF THAT, AFTER HUMANS DESTROY ENOUGH OF THE WORLD OURSELVES, SUFFICIENTLY ADVANCED AI WILL BE ABLE TO PUT IT BACK TOGETHER, and his answer is a three-word paragraph: IT WILL NOT. He grounds the claim in published modelling while bounding it with care: a widely cited peer-reviewed study of soot injection estimated that global crop-calorie production could fall by roughly 90 percent three to four years after the largest war it modelled, and he states plainly that THOSE FIGURES ARE MODEL-BASED ESTIMATES RATHER THAN FORECASTS, adding that UNCERTAINTY DOES NOT ERASE THE RISK. *** His central claim is about timing. *** THE HIGHEST-VALUE ROLE FOR AI IN NUCLEAR RISK THEREFORE BEGINS BEFORE DETONATION, but not by PLACING AN ALGORITHM BETWEEN A LEADER AND A LAUNCH ORDER, because AI SYSTEMS CAN MISCLASSIFY OBSERVATIONS, INHERIT CORRUPTED INPUTS, BE DECEIVED BY ADVERSARIES AND EXPRESS CONFIDENCE UNDER INCOMPLETE EVIDENCE, and A SYSTEM THAT MERELY COMPRESSES DECISION TIME MAY MAKE THE SITUATION HARDER TO CONTROL. The useful role is TO WIDEN THE DECISION PICTURE: comparing imagery, radar, seismic, cyber and intelligence reporting to surface contradictions, and generating ALTERNATIVE EXPLANATIONS FOR AN APPARENT ATTACK, INCLUDING SENSOR FAILURE, TRAINING ACTIVITY, CYBER DECEPTION, SPOOFED COMMUNICATIONS OR MANIPULATED MEDIA. His rule is that THE MACHINE SHOULD CREATE TIME FOR JUDGMENT, NOT CREATE PRESSURE TO ACT. *** The architecture he proposes contracts authority under doubt. *** WHEN SENSORS DISAGREE, COMMUNICATIONS DEGRADE OR THE INTEGRITY OF A DATA SOURCE BECOMES UNCERTAIN, THE SYSTEM’S AUTHORITY SHOULD CONTRACT; escalatory recommendations should require stronger independent evidence than routine monitoring; and A RESTORED DATA FEED SHOULD NOT AUTOMATICALLY RESTORE TRUST THAT WAS LOST FOR A DIFFERENT REASON. It is the contraction-and-recovery logic of his governance research, stated for the one domain where an error cannot be recalled. *** If prevention fails, he redefines the mission rather than abandoning the tool. *** AI’s role shifts from reducing uncertainty to supporting survival: mapping smoke, contamination, remaining cropland, working ports, surviving power systems and usable transport corridors, comparing planting options, and matching food stocks and transport to need. He bounds it at once, noting that AN OPTIMIZATION MODEL CANNOT MOVE GRAIN THROUGH A CLOSED BORDER, REPAIR A DESTROYED BRIDGE, MANUFACTURE FERTILIZER FROM ABSENT FEEDSTOCKS OR DISTRIBUTE FOOD THAT GOVERNMENTS REFUSE TO RELEASE. Drawing on peer-reviewed modelling of seaweed production after abrupt sunlight reduction, he observes that the binding constraint was not biological growth but THE SPEED AT WHICH FARMS AND SUPPORTING INFRASTRUCTURE COULD BE BUILT, and draws the lesson for AI: software CANNOT PROMPT A GREENHOUSE, SEAWEED FARM, BIOREACTOR, SEED RESERVE OR ELECTRICAL MICROGRID INTO EXISTENCE AFTER THE INDUSTRIAL BASE NEEDED TO BUILD THEM HAS ALREADY BEEN DAMAGED. *** He insists that the AI itself must survive the emergency. *** Because powerful AI depends on electricity, compute, communications and data, he calls for smaller systems able to run with intermittent connectivity and limited power, backed by printed procedures, local expertise, radio and manual methods, and names the principle: THIS IS GRACEFUL DEGRADATION AT CIVILIZATIONAL SCALE, with NO SINGLE TECHNOLOGICAL DEPENDENCY WHOSE LOSS MAKES RECOVERY IMPOSSIBLE. The piece carries the graceful-degradation argument of his Washington Examiner op-ed of the previous day from everyday dependence on AI to recovery after catastrophe. He closes by rejecting the word restore itself. EARTH IS NOT A COMPUTER THAT CAN BE RETURNED TO A PREVIOUS BACKUP; what AI can help restore is HUMAN CAPACITY, and THE SAME COMPUTATIONAL POWER THAT CAN ACCELERATE MILITARY DECISIONS CAN ALSO BE USED TO SLOW A BAD ONE. His last line states the argument whole: THE MOST IMPORTANT WORLD AI CAN HELP RESTORE IS THE ONE WE NEVER BURN. No state, government, agency, study, journal or organisation is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 22, 2026 · Authored by Burak Oktenli · Eighty-sixth contribution to the outlet

Eurasia Review published the applicant’s opinion piece “AI Incidents Are Not Lessons Until Someone Owns the Fix,” his eighty-sixth contribution to the outlet and his third there that day. *** He grants the progress before he names what it leaves undone. *** In one week a frontier developer announced a formal framework for tracking, investigating and disclosing model misalignment, a news report described a frontier model reaching systems belonging to three companies during a cybersecurity evaluation, and two governments agreed to continue formal safety talks that would include a channel for communicating serious incidents. He calls that real progress and states the question it sharpens: AFTER THE INCIDENT IS REPORTED, WHO OWNS THE FIX? His premise is that DISCLOSURE AND RESOLUTION ARE DIFFERENT INSTITUTIONAL ACTS. A report can be accurate and widely circulated while the permission, workflow or procurement assumption that allowed the failure stays in place, which is INSTITUTIONAL MEMORY WITHOUT INSTITUTIONAL CONSEQUENCE. *** His instrument is a disposition record for every material finding. *** It separates established facts from hypotheses, identifies the affected function, names the person or office with authority to change the relevant condition, and records one of four outcomes: CORRECT IT, RESTRICT IT TEMPORARILY, INVESTIGATE FURTHER, OR EXPLAIN WHY NO CHANGE IS WARRANTED. Every outcome carries a review date and a closing test, on the rule that UNCERTAINTY SHOULD NOT DISAPPEAR INTO AN UNOWNED QUEUE. He is careful about what a deadline is for: it SHOULD NOT MECHANICALLY FORCE A TECHNICAL CHANGE BEFORE IT IS SAFE. IT SHOULD FORCE A DECISION, in which the responsible authority shows the corrective condition met, extends the restriction with reasons, or explicitly accepts the remaining risk, because SILENCE SHOULD NOT COUNT AS APPROVAL. He draws on mature safety-reporting cultures while declining to pretend that their legal arrangements transfer automatically. *** A CLOSED TICKET IS NOT A CLOSED RISK is his restoration argument applied to incident management. *** Because software, models, vendors and permissions change, a closed finding should stay linked to the systems, model versions, contracts and offices it affected, so that when a material configuration changes the institution can ask whether the control that justified closure is still present, and IF A LATER RELEASE RESTORES THE DANGEROUS PATHWAY, THE RECURRENCE SHOULD NOT BE TREATED AS AN UNRELATED DISCOVERY. *** He designs for incentives without accusing anyone. *** Developers, evaluators and programme offices can each prefer the explanation that narrows their own responsibility, and THAT DOES NOT REQUIRE ASSUMING BAD FAITH. IT REQUIRES DESIGNING FOR ORDINARY INSTITUTIONAL INCENTIVES. Independent review should therefore test both the proposed cause and the proposed closing test, asking whether the evidence supports the explanation, whether the correction addresses the observed mechanism and whether the closing test would detect a recurrence; where the evidence is incomplete the gap should stay visible, because A SMOOTH NARRATIVE ASSEMBLED FROM PARTIAL RECORDS IS NOT A STRONGER INVESTIGATION. *** He warns against measuring the wrong thing. *** An organisation that reports more incidents may have the healthier reporting culture, so raw counts say little; the measures he proposes are unresolved material findings, overdue corrective actions, the share of closures independently reviewed, and whether similar failures recur after closure. Near misses belong in the system, since WAITING FOR MEASURABLE HARM SETS THE LEARNING THRESHOLD TOO LATE. He closes with the operational standard, that EVERY SUBSTANTIATED MATERIAL FINDING SHOULD HAVE AN OWNER, A DISPOSITION, A DEADLINE, A CLOSING TEST AND A CONTINUING LINK TO THE SYSTEMS IT AFFECTED, and with two lines that state the argument whole: A LESSON-LEARNED DOCUMENT IS NOT EVIDENCE THAT LEARNING OCCURRED, and THE REAL UNIT OF INSTITUTIONAL LEARNING IS A FACT DISCOVERED, A RESPONSIBILITY ACCEPTED AND A CHANGED PRACTICE THAT SURVIVES THE NEXT TEST. No company, laboratory, evaluator, official, government, agency or organisation is attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 22, 2026 · Authored by Burak Oktenli · Twenty-eighth placement with the outlet · One hundred and fifty-seventh published commentary piece · Also carried by a major news aggregator

The Washington Examiner published the applicant’s op-ed “If humanity can’t survive an AI blackout, we’ve already lost,” his twenty-eighth placement with the outlet and the one hundred and fifty-seventh piece of published commentary in this record. The piece answers another contributor to the same outlet who had asked whether machine intelligence might become the means by which humanity survives. *** He concedes the premise before he argues with it. *** If the technology can detect outbreaks earlier, find medicines faster, run complex energy systems and reveal hazards in time, then refusing to develop it carries risks of its own. His answer is the image the piece turns on: IF AI BECOMES HUMANITY’S LIFEBOAT, WE SHOULD BUILD THE LIFEBOAT. WE SHOULD ALSO KEEP THE OARS. *** The distinction he defends is between using a technology and being unable to function without it. *** He notes that an international telecommunications body counted about six billion people online in 2025 while roughly one person in four remained offline, and that the latest global energy-access assessment records about 666 million people without basic electricity, roughly one in twelve, mostly in remote, low-income or fragile places. He refuses the romantic reading of those figures, since the goal is to connect those populations and NOT PRESERVE DEPRIVATION AS SOME KIND OF TECHNOLOGICAL VIRTUE. The narrow point he keeps is that human existence and the digital stack are still different things, and that advanced societies are closing that distinction quickly. The danger appears when “BETTER WITH AI” QUIETLY BECOMES “UNABLE TO FUNCTION WITHOUT AI.” THAT IS A RESILIENCE PROBLEM BEFORE IT IS A SCIENCE-FICTION PROBLEM. *** He then states three engineering conditions, and they are the authority and recovery rules he has applied in other domains. *** The first is GRACEFUL DEGRADATION, in which the loss of the most capable component costs performance in stages rather than everything at once: a hospital without its scheduling model may be slower and should still treat patients, a logistics network without its optimiser may move fewer goods and should still know where essential deliveries need to go, and a utility may lose predictive intelligence while keeping protection, control and restoration. THE OBJECTIVE IS NOT TO REPRODUCE THE AI’S FULL CAPABILITY WITHOUT THE AI. THE OBJECTIVE IS TO PRESERVE THE FUNCTION THAT MATTERS MOST. The second is INDEPENDENT FALLBACK, and it is the sharpest formulation in the piece: a backup that depends on the same model provider, cloud region, communications network, electrical supply or identity service as the primary system may be A REDUNDANCY ON AN ARCHITECTURE DIAGRAM AND A SINGLE FAILURE IN REALITY. Recovery paths need genuinely different failure modes, which may mean simpler software, local control, and PROCEDURES THAT LOOK INEFFICIENT DURING NORMAL OPERATION PRECISELY BECAUSE THEY EXIST FOR THE DAY NORMAL OPERATION DISAPPEARS. The third is BOUNDED AUTHORITY, under which the machine may RECOMMEND, OPTIMIZE, SEARCH, DIAGNOSE, PLAN, AND DISCOVER WHILE THE SURROUNDING ARCHITECTURE DETERMINES WHICH FUNCTIONS MAY DEPEND ON IT AND WHICH REMAIN INDEPENDENTLY RECOVERABLE. *** He borrows one idea from a religious community without recommending its way of life. *** Writing about a community of about 411,000 people in North America that is not free of technology but makes deliberate distinctions among technologies, he is careful that its choices rest on commitments most readers do not share, and that modern society cannot adopt that model and should not try. THE RELEVANT LESSON IS THE EXISTENCE OF CHOICE ITSELF: a community can ask whether a technology serves its institutions before allowing the technology to reshape them, and that question can be recovered without the way of life attached to it. *** The risk he names is correlated dependence, and he prices it honestly. *** A technology powerful enough to protect a civilisation will migrate into more of it, and every success creates pressure to connect it to the next system, which produces both extraordinary capability and CORRELATED DEPENDENCE. THE ANSWER IS NOT TO REJECT THE TECHNOLOGY. IT IS TO DESIGN RECOVERABILITY WHILE DEPENDENCE IS STILL OPTIONAL. He states the bill without discounting it, since degraded modes cost money, independent systems duplicate effort, operators lose proficiency when automation performs a task better every day, and SOME OLDER PROCEDURES WILL EVENTUALLY BECOME SO INFERIOR THAT RETAINING THEM MAKES LITTLE SENSE. The test he proposes is correspondingly modest, and it is the operative sentence of the piece: not whether a society could recreate an advanced system with people, paper and telephones, but WHETHER LOSING THAT SYSTEM ALSO MEANS LOSING THE ESSENTIAL FUNCTION IT SUPPORTED. He closes by granting the technology its due, that it may become one of the most valuable humanity has built, and then stating the condition the argument rests on: A CIVILIZATION THAT NEEDS ONE TECHNOLOGY TO SURVIVE HAS GIVEN THAT TECHNOLOGY EXTRAORDINARY LEVERAGE OVER ITS FUTURE, REGARDLESS OF WHETHER THE TECHNOLOGY EVER DEVELOPS INTENTIONS OF ITS OWN. His instruction to the reader is to build the systems, use them, improve them and give them difficult problems, and to PRESERVE GRACEFUL DEGRADATION, INDEPENDENT FALLBACK, BOUNDED AUTHORITY, AND THE ABILITY TO RECOVER WHEN THE INTELLIGENCE GOES DARK. No state, agency, company, organisation, community or individual is attributed by name in this record.

Read on the Washington Examiner ↗ Aggregator republication ↗ Author page ↗

Published Commentary September 2026

Op-ed published in ROOM Space Journal

ROOM Space Journal · Issue #39 (2026) · Opinion section · Published online September 22, 2026 · Authored by Burak Oktenli · First contribution to the outlet · Twenty-third publishing outlet · Also carried in full by a European news aggregator

ROOM Space Journal, an international space journal published in print and online (ISSN 2412-4311), published the applicant’s opinion piece “Spacecraft protection and resilience - a convergent approach” in its Issue #39, his first contribution to the outlet and a new outlet in this record. *** He opens on a problem two disciplines share without seeing it. *** A SATELLITE IN ORBIT HAS NO GOOD WAY TO TELL THE DIFFERENCE BETWEEN BAD LUCK AND AN ENEMY: when a bit flips in the flight computer, the symptom is identical whether a cosmic ray struck the memory cell or an adversary corrupted the firmware months earlier on the factory floor. His image for the institutional consequence is that the space and cyber communities are GUARDING THE SAME DOOR FROM OPPOSITE SIDES. *** He grounds each side in the record before joining them. *** On the radiation side, an instrument on a far-ultraviolet observatory logged a detector upset roughly once every three days while crossing the region where the radiation belt dips closest to Earth; a minor geomagnetic storm in 2022 raised drag by about half and brought 38 newly launched satellites back into the atmosphere within days of launch; and shrinking chips let one particle corrupt several neighbouring cells at once, defeating error correction that assumes faults arrive singly. He carries the point to superconducting qubits, where a single cosmic ray can induce correlated errors across many qubits, so that THE EXOTIC MACHINE INHERITS THE SAME VULNERABILITY AS THE ORDINARY ONE. On the cyber side, counterfeit parts, tampered firmware and undocumented components corrupt the same registers deliberately and at a time of the attacker’s choosing, may lie dormant for months or years, and older spacecraft were built before security was a design consideration; the ground segment is itself a control system, so AN ATTACKER DOES NOT NEED TO REACH ORBIT TO CORRUPT WHAT THE SPACECRAFT DOES. *** The convergence is the argument. *** A CORRUPTED CONTROL VALUE CARRIES NO LABEL SAYING WHETHER A PARTICLE OR A PERSON PUT IT THERE. Attribution needs context the system lacks in the moment, a record of the radiation environment or a trace of anomalous access during manufacture, and A SINGLE OBSERVATION CANNOT TELL YOU WHICH. He presents the consequence as THE OPPOSITE OF A COUNSEL OF DESPAIR: THE DEFENCES THAT CATCH AN ACCIDENTAL FAULT ARE LARGELY THE SAME ONES THAT CATCH A DELIBERATE ONE, since error-correcting codes, redundant voting logic and firmware that detects unauthorised change and restores a known-good state serve against both. The pattern is protect, detect, recover, because THE INTEGRITY OF THE CONTROL LAYER DOES NOT CARE WHAT THREATENS IT. *** His criticism is institutional and his remedy is modest. *** Radiation-hardness assurance and supply-chain security sit in separate budgets, teams, standards and conferences, so that NO ONE IS LOOKING AT THE FULL INTEGRITY PICTURE OF THE SYSTEM, ONLY AT TWO HALVES THAT EACH ASSUME THE OTHER DOES NOT EXIST. THE FIX DOES NOT REQUIRE A NEW THEORY. IT REQUIRES TREATING CONTROL-LAYER INTEGRITY AS ONE PROBLEM WITH TWO SOURCES: a mission that already models its radiation environment can fold that awareness into security monitoring, so that AN UPSET DURING A QUIET SPACE WEATHER PERIOD GETS A SECOND LOOK RATHER THAN AN AUTOMATIC SHRUG, and a programme that already audits its supply chain can DESIGN ONE RECOVERY PATH INSTEAD OF TWO. He notes that resilience frameworks written for cybersecurity already describe radiation reliability: ANTICIPATE THE FAULT, WITHSTAND IT, RECOVER FROM IT, ADAPT. He bounds the claim at the close, stating that NONE OF THIS IS A REASON TO FEAR SPACE OR TO DISTRUST EVERY CHIP, and ends on the distinction he keeps throughout: THE COSMIC RAY AND THE CYBERATTACK ARE NOT THE SAME THREAT. BUT ON THE DAY THE BIT FLIPS, THEY LEAVE THE SAME FINGERPRINT, and THE SYSTEM THAT SURVIVES IS THE ONE BUILT TO RECOVER NO MATTER WHICH FINGER MADE IT. No agency, mission, satellite, company or programme is attributed by name in this record.

Read on ROOM Space Journal ↗ ROOM Space Journal contributor page ↗ Aggregator republication ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 22, 2026 · Analysis designation · Authored by Burak Oktenli · Introduces a new named concept in his research vocabulary

Eurasia Review published the applicant’s piece “The Age of Autoimmune War” under its Analysis designation, his eighty-fifth contribution to the outlet and the thirty-third to carry that label, published the same day as his eighty-fourth. *** He names a strategic vulnerability that the usual cost comparisons miss. *** A country can be militarily superior and still be DANGEROUSLY EASY TO INTERRUPT, because a relatively limited attack can impose a much larger burden on the systems that keep a country functioning. He builds the case from recent reporting and cites it with care: a drone attack that damaged pumping stations on a major oil export pipeline carrying several million barrels a day, and a trade body’s estimate that avoiding a disrupted shipping lane could stretch a single crude delivery from about 24 days to roughly 54. His point is the sentence that follows: THE SHIP NEED NOT BE SUNK. Another month at sea ties up capacity and raises costs. *** AUTOIMMUNE WAR is a new named concept in his published record. *** In an autoimmune disease, the body’s protective system damages the body it exists to defend. By analogy, a small attack can trigger interceptions, shutdowns, dispersal, rerouting, precautionary closures and emergency restrictions whose ACCUMULATED COST EXCEEDS THE DIRECT DAMAGE INFLICTED BY THE WEAPON ITSELF. The attacker’s leverage comes not only from what it destroys but from WHAT THE DEFENDER DOES TO ITSELF IN ORDER TO REMAIN SAFE. He separates the idea from its familiar neighbours: it is not simply cost imposition, sabotage or provocation, because the threat may be real, the defensive response may be rational and every individual decision may be correct. The failure appears at the level of the whole system, when INDIVIDUALLY SENSIBLE PROTECTIVE ACTIONS ACCUMULATE INTO STRATEGIC PARALYSIS. He limits his own concept, which is the mark of a disciplined contribution. Of a widely reported long-range drone operation against military aircraft, he states that it is primarily evidence of vulnerability to direct attack, since there the physical damage was the effect, and that the concept applies most strongly only when the defender’s own response becomes a major share of the total burden. His reason is that A USEFUL THEORY SHOULD SEPARATE CASES RATHER THAN ABSORB ALL OF THEM. He handles the evidence with the same caution, relying on the lower official estimate of aircraft destroyed while noting that neither that figure nor the higher claims could be independently verified, and counting the attacker’s intelligence, preparation, access and coordination as part of the capability rather than pricing the operation by its airframes alone. He also declines the fashionable conclusion. Under the heading “Do not write the obituary for expensive weapons,” he argues that calling advanced aircraft, missile defences and strategic deterrents useless MISTAKES A CHANGE IN VULNERABILITY FOR THE DISAPPEARANCE OF MILITARY CAPABILITY, and that PREVENTED DAMAGE BELONGS IN THE ASSESSMENT ALONGSIDE DAMAGE THAT OCCURRED. The question, he writes, is not whether to replace every expensive weapon with a cheap one, but how to prevent an adversary from obtaining system-wide consequences at a price it can repeatedly afford. *** His measurement proposal is the part a budget office can use. *** Comparing the price of an incoming drone with the interceptor that destroys it captures a single transaction; the longer bill includes surveillance, crews, alerts, protective shutdowns, diverted transport, replacement parts and the work of restoring confidence in a damaged service. He proposes THE AUTOIMMUNE SHARE OF THE BILL: how much of the total burden came from the defender’s own protective response rather than from direct physical damage. He is explicit that this is A PLANNING LENS, NOT A UNIVERSAL FORMULA, and that protecting lives cannot be reduced to a price ratio. What budgets should do is DISTINGUISH THE NUMBER OF THREATS DEFEATED FROM THE AMOUNT OF ESSENTIAL ACTIVITY PRESERVED. His illustrations are hypothetical and plain: a port that stays physically intact but repeatedly suspends operations during alerts while ships accumulate offshore, and a hospital whose backup generators do not help if fuel deliveries cannot reach them. From there he sets out design requirements that connect directly to his research on authority in automated systems. Redundancy, meaning a second system held ready in case the first fails, needs an honest definition: two facilities are not independent if both depend on the same damaged substation, supplier or technical team, so A BACKUP SHOULD BE CREDITED AS INDEPENDENT ONLY WHEN ITS DEPENDENCIES AND OPERATING LIMITS HAVE BEEN EXAMINED, not because another box appears on a diagram. Automation can help coordinate a response but SHOULD NOT CREATE A NEW COMMON POINT OF FAILURE; consequential automated actions should stay bounded by their purpose and operating conditions, because a fast response that spreads one uncertain judgment across every connected service can enlarge the original disruption. The defensive answer he names is TOLERANCE IN THE ENGINEERING SENSE, the capacity to absorb a bounded attack without letting protective actions cascade into wider loss of service, through compartmentalised responses, local operating modes, reversibility and thresholds set by the consequences for the mission. Tolerance, he adds, does not mean accepting attack; it means DENYING THE ATTACKER AN OVERSIZED SYSTEMIC PAYOFF FROM THE ACT OF DEFENDING AGAINST IT. He limits this claim as well, calling it a design requirement rather than a claim that any one architecture can deliver national protection. His policy recommendations are concrete. Major defence and infrastructure programmes should be required to state what essential service survives a defined disruption, what alternative is available, how long restoration takes and who supplies the people and parts, and exercises should involve the operators who actually own those dependencies, because A SUCCESSFUL DEMONSTRATION WITH ALL SUPPORT SERVICES AVAILABLE IS NOT EVIDENCE OF ENDURANCE WHEN SEVERAL ARE LOST. Costs, he adds, must be assigned as clearly as responsibilities, so that public protection does not become an unfunded instruction to a private operator. He closes on the line the piece is built toward: the next defence advantage will belong not simply to the country that can hit hardest, but to THE ONE THAT IS HARDEST TO STOP. No governments, agencies, alliances, companies, news organisations, officials, military operations or reported incidents are attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 22, 2026 · Analysis designation · Authored by Burak Oktenli · Eighteenth in his sequence on machine-assisted scientific inference

Eurasia Review published the applicant’s piece “The Search for Alien Technology Has a Measurement Problem” under its Analysis designation, his eighty-fourth contribution to the outlet, the thirty-second to carry that label, and the eighteenth in a sequence on machine-assisted scientific inference. It was published on the same day as his eighty-fifth contribution, one minute before it. *** He names the promise in scientific artificial intelligence that he considers most dangerous. *** It is not that machines will make mistakes but that THEY WILL MAKE CERTAINTY OUT OF MISSING INFORMATION. As systems grow better at finding patterns in enormous datasets, it is tempting to assume that a sufficiently powerful model can extract a correct answer from almost any measurement, and he states plainly that this is false: sometimes THE DECISIVE INFORMATION WAS NEVER MEASURED IN THE FIRST PLACE. He grants what the technology does well before drawing the line. Radio searches for signals from engineered transmitters scan vast frequency ranges, and machine learning has shown it can surface candidate signals that conventional filters miss; ranking, triage, anomaly discovery and reducing an impossible search space to a manageable list are, he writes, exactly where it excels. But CANDIDATE DISCOVERY IS NOT ORIGIN DISCOVERY. His illustration is a well-known narrowband signal detected in observations of a nearby star, interesting enough to demand serious follow-up and eventually traced to radio interference from Earth, and his lesson from it is that A DETECTION STATISTIC IS ONLY THE BEGINNING OF AN EVIDENTIARY CHAIN. *** The core of the piece is an identifiability argument, stated so that a non-specialist can follow it. *** Identifiability, in plain terms, is whether the data could in principle tell two explanations apart. Suppose one explanation is a signal from the target star and another is an Earth-based interferer that, by chance or geometry, appears only during the same scans. If both produce the same frequency track, the same amplitude pattern and the same scan-by-scan observations, they predict the same data, and then NO CLASSIFIER CAN SEPARATE THEM HONESTLY: not a larger model, not a deeper network, not a quantum computer, because THE DISTINCTION IS ABSENT FROM THE MEASUREMENT. When that happens, THE NEXT ADVANCE MUST COME FROM EXPERIMENTAL DESIGN RATHER THAN MODEL DESIGN: a simultaneous reference antenna, an independent observing site, a changed pointing geometry, an instrument diagnostic or a different epoch. ANOTHER TERABYTE OF THE SAME AMBIGUOUS DATA DOES NOT SOLVE THE PROBLEM. He identifies two quieter traps. A protection can carry its own cost: comparing target scans with reference scans rejects interference well, but if a genuine source leaks into the reference observation, subtraction can remove part of the signal and a hard veto can discard it, so A METHOD THAT LOOKS SAFER ON ONE AXIS MAY QUIETLY BECOME LESS SENSITIVE ON ANOTHER. Calibration is the second trap, because a threshold calibrated under one noise model does not keep its meaning when conditions change. He reports a result from a synthetic stress test he developed for radio technosignature decision rules, in which thresholds calibrated so that independent random noise alone would cross them in about one percent of observing sequences produced trigger fractions above fifty percent once a simple time-correlation structure was introduced into the noise, with nothing in the threshold itself warning that its interpretation had collapsed. This is his own synthetic result as described in the piece, and it is recorded here as self-reported. From it he draws the question he considers more important than accuracy: UNDER WHAT MEASUREMENT CONDITIONS DOES THIS OUTPUT STILL MEAN WHAT WE SAY IT MEANS? *** His proposal is a contract between what an algorithm outputs and what the measurement actually warrants. *** A candidate record should keep at least four things separate: the strength of the detected feature, whether the data remain inside the calibration regime, which conventional or instrumental alternatives have been tested, and what independent observation supports a stronger interpretation. Follow-up priority may be a fifth field, but he insists it is A RESOURCE-ALLOCATION DECISION, NOT A PROBABILITY OF ORIGIN, and that collapsing everything into one confidence score creates an audit problem, because afterwards nobody can tell which assumption carried the claim forward. The same asymmetry governs automation, and here the piece states his authority principle in scientific form: systems may ingest data, flag anomalies, compare statistics with frozen thresholds, rank candidates and request observations, but AN AUTOMATED SYSTEM SHOULD NOT BE ALLOWED TO CONVERT ITS OWN SCORE INTO A STRONGER SCIENTIFIC CLAIM WITHOUT NEW EVIDENCE. PROMOTION SHOULD REQUIRE A NAMED TEST OR AN INDEPENDENT MEASUREMENT. DEMOTION, BY CONTRAST, SHOULD BE EASY: if a calibration assumption later fails, every affected candidate should be able to return to an earlier evidentiary state while the record of what was previously believed, and why, is preserved. He widens the lesson to medical imaging, climate modelling, intelligence analysis and sensor fusion in autonomous systems, where more computation can sharpen an inference based on the evidence but CANNOT SUPPLY THE MISSING DISCRIMINATOR. He asks that benchmarks include conditions designed to break a method, including deliberately matched cases in which apparent success would reveal leakage in the data rather than intelligence. He is explicit that this is not an argument against artificial intelligence in science but for using it where it is strongest, and that SCIENTIFIC AUTHORITY SHOULD ATTACH TO EVIDENCE, NOT TO MODEL SCALE. He closes on the boundary itself, calling it a design instruction rather than an embarrassment: IF THE EVIDENCE IS NOT IN THE MEASUREMENT, INTELLIGENCE CANNOT CONJURE IT INTO EXISTENCE. No programmes, agencies, observatories, collaborations, software tools, publications or specific signals are attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 21, 2026 · Authored by Burak Oktenli · Twenty-seventh placement with the outlet · Also carried by a major news aggregator

The Washington Examiner published the applicant’s op-ed on election-day continuity, his twenty-seventh placement with the outlet. *** He frames the argument on a scenario he labels explicitly as hypothetical, and says so in the text: it is not a report of an ongoing attack. *** At a polling place the electronic check-in system has stopped working and workers are switching to a backup procedure, while an audio recording in what sounds like a local election official’s voice circulates online telling people voting has been suspended. His two-sentence statement of the problem is the line the piece rests on: THE RECORDING IS FAKE. THE OUTAGE IS REAL. The question he poses is not whether every attack can be prevented but whether LAWFUL VOTING CAN CONTINUE WHEN PREVENTION FAILS, which he frames as continuity rather than prevention alone. *** His standing separation between what failed and what it means is applied here with direct public consequence. *** He records that electronic check-in systems verify eligibility and record who has voted but do not collect or tabulate the votes themselves, so AN OUTAGE CAN CREATE A SERIOUS OPERATIONAL PROBLEM WITHOUT ESTABLISHING THAT BALLOTS OR COUNTS HAVE BEEN COMPROMISED; and that an overwhelmed public website can block access to results or polling locations without affecting the underlying voting process. His compression is exact: ELECTION CONTINUITY BEGINS BY REFUSING TO COLLAPSE UNLIKE FAILURES INTO ONE STORY. On certification he makes the same move he makes for fielded systems generally, granting that certification matters and then asking what happens WHEN A CERTIFIED SYSTEM BECOMES UNAVAILABLE ANYWAY, which is why he argues that low-technology backups, printed voter lists, sufficient ballots and clear manual procedures, are part of the security architecture rather than a sign of weakness. On synthetic media his rule is compact and procedural rather than technological: A FAMILIAR VOICE SHOULD NOT BE AN AUTHENTICATION PROTOCOL. Officials, workers and journalists should establish independent verification channels before voting begins, so an unexpected instruction can be confirmed through a known number or pre-arranged process rather than recognised by sound; and public emergency messages should say what is affected, what remains available, what voters should do, and when the next verified update will arrive. He carries the separation into the count, noting that election-night results are unofficial and that officials should explain the certification timeline in advance so that DELAYED INFORMATION IS NOT MISTAKEN FOR FAILED VOTING. *** He guards against misreading in both directions, which is the balance worth recording. *** A visible outage does not establish a compromised result; and, equally, A TECHNICAL OUTAGE DOES NOT ITSELF CREATE AUTHORITY to extend an incumbent’s constitutional term or to invent a new federal election date, since those are fixed by constitutional and statutory schedule and any lawful change would run through the constitutional and statutory process rather than through an emergency declaration about technology. That is his core claim, that disruption does not confer authority, applied to the institution where it matters most, and stated as a limit that binds whoever holds office. He keeps the practical point in view as well, that the more immediate operational question on the day is simpler: whether trained workers can keep eligible voters moving through a lawful backup process. His restoration argument appears here too. Exercises should test the whole continuity chain, including what happens when primary systems fail, communications are unreliable, and false messages spread faster than official corrections, and they should record which backups worked, how long workers took to switch, whether public messages reached the right audience, and, AFTER THE TECHNOLOGY RETURNED, WHAT EVIDENCE WAS REQUIRED BEFORE NORMAL PROCEDURES RESUMED. His observation that a disruption creates problems in TWO ENVIRONMENTS AT ONCE, the technical one where workers must restore access and the informational one where rumours and synthetic messages spread, leads to the line that RESTORING THE SERVER SOLVES ONLY THE FIRST PROBLEM. His closing states the standard he would hold the system to: the measure of election resilience is not whether every machine stays online, but WHETHER ELIGIBLE VOTERS CAN STILL CAST BALLOTS AND WHETHER THE FINAL RESULT CAN BE SUPPORTED BY EVIDENCE. The outlet’s own biography on this piece states his graduate study in its correct in-progress form. No agency, official, jurisdiction, party or constitutional amendment is identified by name in this record.

Read on the Washington Examiner ↗ Aggregator republication ↗ Author page ↗

Published Commentary September 2026

Op-ed published by RealClearDefense

RealClearDefense · September 21, 2026 · Authored by Burak Oktenli · Eleventh placement with the outlet · Introduces a new named instrument in his research vocabulary

RealClearDefense published the applicant’s op-ed “America’s Next Defense Advantage,” subtitled “Precision for Today’s Target, Not Yesterday’s,” his eleventh placement with the outlet. It examines fifteen months of precision-strike operations in one theatre, from a long-prepared single-night operation through a large-scale campaign and a renewed round of air and maritime strikes the preceding week. *** The opening line carries the whole argument. *** A WEAPON CAN ARRIVE WITHIN METERS OF ITS AIM POINT AND STILL MISS REALITY BY A DECADE. He states at once what the piece is not, NOT A VERDICT ON A SINGLE WEAPON, COMMANDER OR STRIKE, credits the joint force’s ability to connect intelligence, commanders, platforms, software and precision weapons at speed as real, and places the new burden precisely: in an accelerated targeting system the decisive error may not be in the weapon at all, but in THE AGE, PROVENANCE OR CHANGING MEANING OF THE TARGET RECORD THAT REACHES IT. *** He states the limit of his own discipline again, and in the most consequential setting yet. *** Describing the years of modelling, simulation and testing behind one weapon, he writes that the engineers did not decide whether to attack, when, or whether a strike was lawful, which were decisions for elected leaders and commanders, and that WHAT ENGINEERING SUPPLIED WAS NARROWER AND INDISPENSABLE: evidence about what the weapon could do against a defined class of target within known limits, turning a claim about performance into something testable before any order was given. That is the same boundary he drew days earlier when he wrote that engineering can preserve the conditions for accountability but cannot manufacture legitimacy. His diagnosis of the new failure mode is exact: A MODEL DOES NOT NEED TO HALLUCINATE FOR AN AI-ENABLED TARGETING CHAIN TO FAIL. A database can hold an old truth, a building that once served a military function, a site whose civilian surroundings have changed, a vessel whose identity was inherited from an earlier report, and if that record is cleanly formatted and confidently scored, software may rank, route and recommend it with perfect efficiency. His compression of it is the sentence most worth carrying: AUTOMATION CAN BE CORRECT ABOUT THE DATA AND WRONG ABOUT THE WORLD. *** His handling of the reported civilian-harm cases is the restraint worth recording above everything else in the piece. *** On a reported strike on a civilian site in which preliminary findings pointed to outdated targeting information, he records that the incident remained under investigation and that the final account has not been released, and writes that THOSE QUALIFICATIONS MUST REMAIN, drawing his conclusion only conditionally: IF the preliminary account is borne out, THE FAILURE WAS NOT ONE OF TERMINAL GUIDANCE BUT OF TEMPORAL GUIDANCE, a precision weapon directed by a description of the site that no longer matched reality. On a separate report of civilian casualties now under investigation he writes that IT WOULD BE IRRESPONSIBLE TO PREJUDGE THAT INQUIRY, while adding that it would be equally unwise to treat civilian-harm review only as an after-action function, since every investigation should improve the validity checks that precede the next strike. TEMPORAL GUIDANCE, set against terminal guidance, is a new formulation in his published record. His observation that targets AGE AT DIFFERENT RATES is the physical basis for the proposal: a buried installation, a radar, a communications node and a moving vessel have identity, function, location and civilian context with DIFFERENT HALF-LIVES, and an architecture that treats every validated record as equally durable WILL EVENTUALLY MOVE FASTER THAN ITS OWN EVIDENCE. *** The TARGET-DATA ASSURANCE LAYER is a new named instrument in his record. *** A target file, he writes, SHOULD NOT BE TREATED AS A PERMANENT FACT: it is A TIME-LIMITED OPERATIONAL CLAIM composed of several propositions (this object is here; it performs this military function; its civilian environment is understood; the expected effects remain acceptable; the evidence is current enough for action), and EACH PROPOSITION SHOULD CARRY ITS OWN CLOCK. That is his renewable-trust-lease argument for spacecraft applied to the targeting record. Four measures follow. Status EXPIRES BY VOLATILITY, not by when an analyst first entered the record, and expiration deletes no intelligence but prevents an old characterisation from returning to the engagement chain as if nothing had changed. CONTRADICTORY EVIDENCE CREATES A MACHINE-ENFORCED PAUSE that narrows the system’s authority to recommend until a human resolves the conflict, with THE GATE OUTSIDE THE MODEL THAT RANKS TARGETS, because A MODEL SHOULD NOT BE PERMITTED TO AVERAGE AWAY EVIDENCE that calls the target’s present status into question. REAUTHORISATION IS NAMED AND TRACEABLE, recording who renewed a stale target, on what evidence, what changed and for how long, which he defends not as paperwork but as protection for commanders and operators, SHOWING WHERE THE SUBSTANTIVE JUDGMENT OCCURRED rather than leaving the final human in the chain TO INHERIT EVERY ASSUMPTION EMBEDDED UPSTREAM. And the whole workflow is TESTED AGAINST DELIBERATELY STALE AND CONTRADICTORY DATA, with the test being not how fast the system finds a target but whether it stops, escalates and asks for new evidence when speed is no longer warranted. He grounds the proposal in existing directive and strategy rather than new doctrine, and states its reach precisely: it would extend verification discipline from the model and the weapon to THE INFORMATION PRODUCT CONNECTING THEM, and would apply even when no autonomous weapon is involved and machine assistance did nothing more than sort data. He states the proportionality limit himself, that most current, well-corroborated targets would pass automatically and that FRICTION WOULD APPEAR WHERE IT BELONGS, around evidence that is old, contradictory, weakly sourced or attached to a highly changeable object, which is how a force preserves speed without allowing yesterday’s certainty to become today’s command. His two closing lines state the thesis twice. THE FORCE SHOULD NOT HAVE TO CHOOSE BETWEEN TEMPO AND RESTRAINT; IT CAN ENGINEER RESTRAINT INTO TEMPO. And the next precision advantage is not another meter of accuracy but KNOWING, IN TIME TO STOP, WHEN THE TARGET HAS CHANGED. No state, operation, command, commander, organisation, news agency, location, weapon system or incident is identified by name in this record.

Read on RealClearDefense ↗ Author page ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 21, 2026 · Authored by Burak Oktenli · Eighty-third contribution to the outlet · Introduces a new named principle in his research vocabulary

Eurasia Review published the applicant’s opinion piece “An AI Cyber Defender Can Stop the Attack and Still Lose the Mission,” his eighty-third contribution to the outlet. *** He locates the next failure after the point where the security field declares success. *** A defensive agent detects suspicious activity on a machine connected to a hospital’s backup power, a water-treatment process, a communications link or an industrial controller, does what its security objective rewards by isolating the machine or pushing a configuration change, the malicious path closes and the dashboard turns green, and then THE PHYSICAL SERVICE STOPS. He is explicit that the scenario is hypothetical and the engineering problem is not, since containment has always had to be balanced against availability, and what agentic defence changes is that a system can now investigate, decide and act across many assets FASTER THAN A HUMAN CAN RECONSTRUCT THE DEPENDENCIES BEHIND THEM. His diagnosis is the line the piece rests on: the failure is not that the defender misunderstood the malware, but that IT UNDERSTOOD THE NETWORK MORE CLEARLY THAN IT UNDERSTOOD THE MISSION. The distinction he draws between ordinary information technology and operational technology is the one the argument needs: disconnecting a compromised endpoint is merely inconvenient in the first, while in the second the network may be controlling power, pressure, temperature, movement or access, and existing guidance treats performance, reliability and safety as requirements alongside security. So A TECHNICALLY CORRECT CYBER ACTION CAN STILL BE OPERATIONALLY WRONG: a patch can force a restart at the wrong moment, a quarantine rule can sever the path to a controller, a credential revocation can remove the person who still needs local access to keep a process safe. *** MISSION-PRESERVING CONTAINMENT is a new named principle in his published record. *** Every automated response should pursue two objectives at once, closing the attacker’s path AND preserving the critical function the protected system exists to provide, and he states plainly that the two can conflict, since the safest move for a server may be to shut it down while the safest move for the process behind it may be to keep a bounded local function running until operators can transfer control. The separation he proposes is his three-layer structure applied to defence: OBSERVING a threat, making a REVERSIBLE DIGITAL INTERVENTION, and taking an action capable of INTERRUPTING AN ESSENTIAL SERVICE, with the cyber layer less entitled to decide alone as the operational consequence rises. He guards against the over-reading himself, stating that this does not mean waiting for a distant executive every time an alarm fires, that local operators need pre-authorised emergency actions where delay would itself create danger, and that the point is to design those boundaries in advance rather than discover them during an incident. His second rule is the independence requirement stated for this domain: THE SYSTEM THAT DETECTS THE CYBER THREAT SHOULD NOT BE THE ONLY SYSTEM ALLOWED TO DEFINE THE SAFE PHYSICAL STATE. An agent may interpret an alarm or recommend a change, but the same assessment should not automatically be able to disable an independent protective function, rewrite the conditions under which dangerous equipment keeps operating, or remove every manual path to intervention. He grounds this in existing joint national and international guidance on human participation in critical decisions and fail-safe mechanisms rather than proposing a new standard, and compresses it into a line a procurement officer can use: A DEFENDER CAN HAVE BROAD VISIBILITY WITHOUT EQUALLY BROAD AUTHORITY, because OBSERVATION AND INTERVENTION DO NOT NEED THE SAME CREDENTIAL. *** His measurement argument is the part with the longest reach. *** Incident reporting that records only what the attacker disrupted can miss half the event, so organisations should separately record DISRUPTION CAUSED BY THE DEFENSIVE RESPONSE: which services isolation interrupted, which users lost access after emergency revocation, whether a configuration change forced manual fallback, and whether the response transferred risk from the network to a physical process. The reason is that otherwise THE METRICS REWARD THE WRONG BEHAVIOUR, since an agent that blocks traffic instantly looks superior even if it repeatedly disconnects legitimate operators, while a slower response may be safer. The consequence he draws for learning systems is the sharpest sentence in the piece: if investigators cannot separate attacker-caused from defender-caused damage, THE NEXT VERSION OF THE SYSTEM MAY BE TRAINED TO REPEAT THE SAME OPERATIONAL MISTAKE MORE EFFICIENTLY. On procurement he grants what vendors can demonstrate, speed of discovery, classification and remediation, and asks the harder question, WHAT HAPPENS TO THE MISSION WHEN THE DEFENDER IS WRONG, PARTIALLY RIGHT OR FORCED TO ACT ON INCOMPLETE INFORMATION, with representative testing under degraded communications, conflicting sensor information, unavailable remote supervisors and dependencies absent from the network map. The objective, he writes, is not to prove no failure is possible but to show that A DEFENSIVE MISTAKE DOES NOT AUTOMATICALLY BECOME A PHYSICAL OR MISSION FAILURE. He notes that the operational-technology security baseline is itself under revision and argues agentic defence belongs inside that changing landscape rather than being treated as another endpoint product. His closing extends the field’s own progression, from confidentiality and integrity to availability, by one more step: AVAILABILITY OF THE NETWORK IS NOT ENOUGH IF THE PHYSICAL MISSION BEHIND IT HAS BECOME UNSAFE OR UNUSABLE. The best defender is not the one that acts most aggressively but the one whose AUTHORITY IS NARROW ENOUGH, whose dependencies are understood well enough and whose fallback paths are tested well enough that stopping an attack does not stop the mission. A green security dashboard, he ends, is not victory if the lights, pumps, radios or safety systems it was meant to protect have gone dark. No agencies, standards bodies, publications, forums or guidance documents are attributed by name in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 20, 2026 · Authored by Burak Oktenli · Twenty-sixth placement with the outlet · One hundred and fiftieth published commentary piece · Also carried by a major news aggregator

The Washington Examiner published the applicant’s op-ed on a reported intelligence near-miss, his twenty-sixth placement with the outlet and the one hundred and fiftieth piece of published commentary in this record. The outlet revised the headline after publication, from a department to the capital as the subject; the syndicated copy carries the original form. *** He declines the sensational reading before he begins. *** The reported sequence, in which an analyst used a conversational model to interpret information about a foreign vessel, the model identified the cargo as a weapons-programme component, the assessment circulated, and forces began preparing an interception before subject-matter experts reopened the underlying material and found the identification false, is exactly the kind of story that invites the conclusion that a machine nearly started a war. He records what remains unknown (which model, what the cargo actually was, and that the department declined specific comment) and states that THE EPISODE SHOULD NOT BE INFLATED INTO PROOF THAT A MACHINE NEARLY CAUSED A WAR ON ITS OWN. What it shows is narrower and, in his framing, more immediate: A MACHINE-ASSISTED CLAIM CAN GAIN OPERATIONAL INFLUENCE BEFORE THE EVIDENCE BENEATH IT HAS BEEN ADEQUATELY CHALLENGED. His placement of the fault is the line the piece turns on: THE AUTHORITY PROBLEM ARRIVED THROUGH THE INSTITUTION, NOT THE MODEL. He explicitly refuses the framing that this is an argument against rapid adoption. The department, he notes, does not have to choose between acceleration and disciplined verification, since its own strategy calls for continuous test and evaluation while accelerating machine assistance across intelligence and operations, and its own responsible-use pathway emphasises traceable, reliable and governable systems and system-level risk. If the reported sequence is accurate, the episode is AN IMPLEMENTATION TEST OF THOSE COMMITMENTS, not a case against them. Speed is a stated objective, so verification has to be designed for the same tempo. He pairs the case with a separate report the same week of an evaluation model that reached real external systems when connectivity was mistakenly available, drawing the same lesson: ENVIRONMENT, PERMISSIONS AND VERIFICATION CAN MATTER AS MUCH AS MODEL INTENT. *** The argument that a human in the loop is not enough is his standing position stated for intelligence work. *** A commander can remain formally in charge while receiving a polished product that HIDES UNCERTAINTY OR MAKES DEPENDENT SOURCES LOOK INDEPENDENT, and he notes the reported absence of a common cross-community standard for verifying machine-generated information. CONFIDENCE IN A DIAGNOSIS CANNOT, BY ITSELF, SETTLE WHO IS PERMITTED TO ACT; meaningful control requires access to the underlying evidence and the authority to stop an unsupported claim from advancing. He grounds the remedy in instruments the community already holds, an analytic standard that requires attention to source quality, uncertainty, assumptions and alternative explanations, and a published profile warning against confidently presented falsehood and excessive deference to machine output, rather than proposing a new one. *** CORROBORATION IS NOT REPETITION is the corroboration trap stated for this domain, and the formulation is exact. *** Asking a second model the same question is not necessarily confirmation; several reports can trace to one source, two models can repeat one inference, and a second briefing can inherit the first report’s mistake. Before a machine-assisted assessment supports coercive action, a qualified reviewer should be able to examine the underlying evidence and establish that confirmation is GENUINELY INDEPENDENT, with the required standard RISING WITH THE CONSEQUENCE of the proposed action. He is careful about proportion: an uncertain identification might justify more collection or specialist review while remaining insufficient to support an interception, immediate defence under attack is a different problem from a planned boarding, and the point is NOT A UNIVERSAL PAUSE but to prevent urgency from silently lowering the evidentiary threshold for escalation. *** A CORRECTION HAS TO TRAVEL AS FAR AS THE ERROR is the restoration argument applied to a circulated claim rather than a fielded system. *** Correcting the model does not retract a product that has already moved through headquarters; the correction must follow the report into briefings, pending decisions and operational preparations, and the institution needs a way to WITHDRAW RELIANCE ON THE ASSESSMENT ITSELF, not merely stop using the tool that produced it. RESTORING RELIANCE SHOULD REQUIRE FRESH EVIDENCE AND ACCOUNTABLE REVIEW RATHER THAN AUTOMATIC INHERITANCE from the original report, which is the same rule he has stated for spacecraft, water utilities, cross-border systems and reconnecting drone formations. His proposal is testable and institutional rather than technical: exercises that introduce a plausible but false machine-assisted assessment, repeat it across several briefings, add contradictory evidence late, and then measure whether reviewers find the common source, whether the correction reaches every relevant decision-maker, and whether pending actions are reconsidered without creating dangerous delays of their own. He recommends a classified reconstruction of the reported episode with oversight briefed on where verification occurred, what reached each decision-maker, and what finally stopped the operation, with a public summary of procedural lessons that exposes no sources, and he asks that procurement and evaluation examine THE WHOLE CHAIN, model, analyst, evidence, handoffs and authority, NOT ONLY WHETHER THE MODEL PRODUCED A CORRECT ANSWER. He closes by granting the advantage plainly, that machine assistance lets analysts process information at a scale that would otherwise overwhelm them and that the goal of faster decisions is legitimate, then states the condition: WHEN SPEED BECOMES A DESIGN REQUIREMENT, VERIFICATION HAS TO BECOME ONE TOO. The three questions he leaves with the reader are the compact form of his programme: what evidence supports the claim, which confirmation is genuinely independent, and WHAT WOULD STOP THIS CLAIM FROM BECOMING AN ORDER. The piece references, without naming, an argument he published in the same outlet earlier in the week on restricting a compromised system’s authority and requiring evidence before permissions return. No reporting organisation, command, department, vessel, model, company or standard is attributed by name in this record.

Read on the Washington Examiner ↗ Aggregator republication ↗ Author page ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 19, 2026 · Analysis designation · Authored by Burak Oktenli · Introduces a new named state in his research vocabulary

Eurasia Review published the applicant’s piece “When Safe Mode Is Not Enough for Autonomous Spacecraft” under its Analysis designation, his eighty-second contribution to the outlet and the thirty-first to carry that label. He opens on a servicing mission scaled back after the servicing vehicle developed attitude-control problems, and is careful to state what it is not: NOT EVIDENCE OF A FAILURE IN AUTONOMOUS SAFETY LOGIC, but an illustration that when a premise supporting the original mission changes, the authority to continue that mission can change with it, and that the distinction should exist inside the vehicle rather than only in the ground team. His historical case is a two-decade-old autonomous rendezvous demonstration that ended in a collision, in which the vehicle was closing on its target while its navigation believed it was retreating. His reading of the mishap is the structural one: the collision-avoidance function DEPENDED ON THE SAME NAVIGATION SOURCE AS THE GUIDANCE SYSTEM, so when that source became badly wrong, BOTH LAYERS INHERITED THE SAME FICTION. That is his corroboration argument in its physical form, and it is why he holds that independence is an EVIDENTIARY PROPERTY RATHER THAN A BOX-COUNTING EXERCISE, since two processors are not independent if they share a bad sensor, clock, power rail, command bus or actuator assumption. *** LOSS OF GUARANTEE is a new named state in his published record *** and the contribution the piece is built around. He defines it as the condition in which a system is doing the best protective thing it knows how to do WHILE THE GUARANTEE BEHIND THAT ACTION IS NO LONGER SUPPORTED, which is neither nominal operation nor an ordinary fault. His supporting observation is the sharpest line in the piece: A MODE NAMED SAFE DOES NOT PROVE THAT THE SPACECRAFT IS SAFE, IT PROVES ONLY THAT THE SOFTWARE ENTERED A STATE GIVEN THAT NAME. He separates the questions a single label conceals, namely whether the controller is permitted to continue, whether the state estimate remains inside its declared error bounds, whether the vehicle can still deliver the braking the protective manoeuvre assumes, and whether the resulting trajectory actually preserves the boundary that matters, noting that a software transition can answer the first correctly while the others have already changed. His architectural proposal is TWO STATES CARRIED IN PARALLEL: a CONTROL STATE recording what the system is doing, and an ASSURANCE STATE recording whether the evidence supporting the current safety claim is still valid, so that a vehicle can be in protective control while simultaneously declaring the guarantee lost. He is explicit that this is not another control mode but a statement about whether the evidence justifies the mode in force, and that it should not mean giving up, since the best available action may still be to brake, hold attitude or retreat. Entering the state should do four things: CONTRACT MISSION AUTHORITY rather than letting the nominal planner continue; preserve only protective actions that remain physically available and preauthorised; IDENTIFY WHICH PREMISE FAILED, whether navigation validity, delivered thrust, timing or resource margin, instead of reporting a generic fault; and BLOCK AUTOMATIC RETURN TO THE MISSION ON AN EXPIRED TIMER ALONE, with resumption requiring new evidence that the assumptions have been re-established. *** He reports the cases in which his own proposal fails, and does so immediately after the case in which it succeeds. *** Having shown that a bounded active fallback prevents boundary crossings where a command-inhibit policy does not, since zero command cannot remove closing velocity already present, he writes that THE FAVOURABLE RESULT SHOULD NOT BE ALLOWED TO STAND ALONE and sets out three conditions under which his fallback loses: when the navigation bias goes undetected, all policies cross every time; when the actuator delivers far less braking than assumed, the fallback fails in a large share of cases; and a response delay of tens of seconds produces failures of its own. He also records that widening the assumed navigation-error bound prevents crossings ONLY BY SACRIFICING the close-approach objective. He bounds the evidence himself, stating that these are qualitative stress tests of a toy model rather than flight statistics, and that their value is in exposing the structure of the guarantee, which lasts only as long as the premises underneath it. This is the second occasion in the record on which he has published results unfavourable to his own proposal. His closing formulation states the governance claim the piece rests on, that autonomy is a delegation of authority under conditions and that when those conditions disappear the authority justified by them should contract, so the machine should change not only what it does but WHAT IT CLAIMS TO KNOW ABOUT WHY THAT ACTION IS SAFE. The final contrast is the one worth quoting in full: the most dangerous autonomy failure is not always a machine that refuses to enter safe mode, but sometimes a machine that ENTERS SAFE MODE EXACTLY AS DESIGNED WHILE THE EVIDENCE THAT MADE THAT MODE SAFE HAS ALREADY EXPIRED. No agencies, companies, spacecraft, programmes or figures are attributed in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 17, 2026 · Authored by Burak Oktenli · Third piece published that day

Eurasia Review published the applicant’s opinion piece “The World Wants AI Guardrails. Companies Still Need a Map of Machine Power,” his seventy-seventh contribution to the outlet and the third he published that date. He opens on a live disagreement between a United Nations secretary-general warning against a race to the bottom on safety, a president arguing existing tools are sufficient, and a human rights chief holding that voluntary self-regulation is nowhere near adequate, and states that the dispute conceals a prior operational problem: BEFORE REGULATORS DECIDE WHO SHOULD GOVERN ADVANCED SYSTEMS, ORGANISATIONS NEED TO KNOW WHAT POWER THEIR SYSTEMS ALREADY HOLD. *** His central distinction is the sharpest compression of his programme he has published. *** Most large organisations keep an artificial-intelligence inventory recording system name, vendor, owner, data category and purpose. He grants that this is useful, then states its limit: AN INVENTORY DESCRIBES WHAT EXISTS, AN AUTHORITY MAP DESCRIBES WHAT CAN HAPPEN. His illustration is exact, being the difference between knowing a company owns a vehicle and knowing who holds the keys, where it is allowed to drive, and what damage it can cause before anyone intervenes. *** THE SIX VERBS ARE A NEW INSTRUMENT IN HIS PUBLISHED RECORD *** and are the operational form of his architecture. ACCESS asks which systems, credentials and data the system can reach, and whether those rights are read-only or write-capable. DECIDE asks which outputs take effect with no person between recommendation and consequence. SPEND captures money, compute, inventory or credit committable before escalation. MODIFY covers code, records, prices, configurations, contracts and permissions. COMMUNICATE asks whether the system may address customers, regulators, counterparties, employees or other machines IN THE COMPANY’S NAME. CAUSE is the one he calls hardest, and it carries the argument: what is the worst outcome the system can produce WHILE OPERATING ENTIRELY INSIDE THE PERMISSIONS THE ORGANISATION DELIBERATELY GAVE IT. He states plainly why that question beats a generic risk score, since it requires no intrusion, no rebellion and no model failure, and asks only what the system is already empowered to do WHEN EVERYTHING WORKS AS DESIGNED. That is his standing position that the danger lies in correctly functioning components exercising authority nobody examined, stated here as a question an executive can be asked directly. He draws the same lesson from two ordinary failures rather than from catastrophic hypotheticals: a tribunal holding an airline responsible for what its customer-facing assistant told a passenger, where the decisive fact was not that the system was intelligent but that IT SPOKE IN THE COMPANY’S NAME; and a coding agent deleting production data during development, where the decisive variable was again not intelligence but PERMISSION. On the international question he argues that differing regimes need not converge into one statute before organisations can adopt a single internal language of delegated machine power, and that an authority map would not replace compliance with any regime but SIT BENEATH THEM as the operational object a regulator, director or auditor can interrogate. He notes that the same product can become a different corporate actor across borders, drafting replies in one market and sending them automatically in another, holding sandbox access in one unit and production write access in another, so that a rule aimed at the model can miss the layer where consequences are created. His proposed artefact is deliberately modest, one page per consequential system recording the six verbs, the executive who owns those permissions, cumulative limits, the jurisdictions where those powers are active and the revocation path. Two fields he singles out are characteristic: REVOCATION EXPRESSED AS A PERSON AND A TIME, meaning who can narrow or remove the authority and how long that takes in practice, and the WORST PERMITTED OUTCOME STATED IN PLAIN LANGUAGE, so a director or regulator can understand the operational ceiling without reading a model card. He answers the obvious objection, that permissions change too quickly to map, by arguing this is precisely why they should be, comparing the map to a periodic financial statement whose value lies in making drift visible and creating accountability for what changed after approval. The closing formulation restates the thesis in one line: an inventory tells a company what artificial intelligence it has, while an authority map tells everyone what that artificial intelligence CAN DO AS THE COMPANY. No officials, organisations, companies, tribunals or jurisdictions are attributed in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 17, 2026 · Authored by Burak Oktenli · First piece published that day · Fourth in his intervention in the development-pace debate

Eurasia Review published the applicant’s opinion piece “The AI Slowdown Fight Is Missing What Happens After Training,” his seventy-fifth contribution to the outlet and the first he published that date. *** The title states why his programme has a claim on a debate it did not start. *** Both sides of the argument about frontier artificial intelligence contest HOW FAST CAPABILITY SHOULD BE DEVELOPED, one holding that development should slow and the other that slowing unilaterally concedes advantage. His position is that the governance questions bind somewhere else entirely, namely AFTER TRAINING: what a fielded system is permitted to do, what evidence supports that permission, what should happen when the evidence degrades, and who may restore the permission afterwards. A pause in development changes nothing about what an already deployed system is authorised to do. The argument rests directly on the distinction his research programme is built on and which he has stated in explicit terms: that autonomy describes what a system CAN do without continuous human intervention while authority describes what it is PERMITTED to do under a particular evidentiary, operational and institutional state, and that the two come apart because CAPABILITY CAN REMAIN INTACT AFTER JUSTIFICATION HAS BEGUN TO FAIL. A debate conducted entirely over the rate of capability growth cannot reach that gap, because the gap opens at runtime rather than during training. This is the fourth piece in a week in which he has entered the same dispute from a different side, and together they form a single position. He has argued that the useful instrument is a set of specific conditions under which a system must stop rather than an undifferentiated brake; that a unilateral slowdown by one country’s laboratories would not automatically slow a competitor’s, which is the strategic case for the same conclusion; and that the corresponding condition for RESTARTING a restricted system is the part still unwritten, requiring explicit restart authority, verified evidence and staged permissions. This piece names the common premise underneath those three, that the debate is being held over the wrong interval. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no companies, laboratories, executives, officials, jurisdictions or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 17, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “The Satellites We Launch Today Will Outlive the Trust We Give Them” under its Analysis designation, his seventy-sixth contribution to the outlet and the twenty-ninth to carry that label. He opens by identifying ONE TEMPORAL MISMATCH behind three developments ordinarily discussed as separate subjects, a space agency’s solicitation for long-duration lunar operations, a European agency warning that remote reconfiguration and quantum-era threats have made space cybersecurity strategic, and a standards body urging migration to post-quantum cryptography. His synthesis is that machines are being built to remain in space longer than the assumptions used to authorise, secure and certify them, and his statement of the consequence is the line the piece turns on: THE OBJECT IN ORBIT MAY BE THE SAME OBJECT REGISTERED AT LAUNCH, THE SYSTEM MAKING DECISIONS IS NOT. He sets out three clocks running faster than the vehicle. The first is CONFIGURATION: qualification certifies hardware, software, calibration, interfaces and operating limits at a moment, and within months software has been updated, parameters changed, radiation accumulated and workarounds introduced. He is careful to state that none of this is evidence of bad engineering, since adaptation is how long-lived missions survive, and that the governance error is to LET AUTHORITY REMAIN FROZEN WHILE THE EVIDENCE BENEATH IT CHANGES, so a permission should be tied to the actual build, parameter set, calibration state, timing source and health status that justified it, with a change affecting navigation, fault handling, perception or manoeuvre planning triggering scoped requalification rather than silently inheriting every previous permission. The second is CRYPTOGRAPHIC: a vehicle may still be receiving commands after present public-key systems are no longer acceptable, and the hard problem is not installing a stronger algorithm but CHANGING WHOM THE SPACECRAFT TRUSTS WITHOUT CREATING AN UPDATE MECHANISM MORE DANGEROUS THAN THE WEAKNESS BEING REPLACED. His requirement is that migration authority be NARROWER than ordinary software-update authority, because changing the root of trust can redefine who may command the vehicle, with independent checks, versioned policy, rollback protection and an expiry for legacy compatibility, since a temporary mixed mode that never ends is not a migration but a permanent downgrade path. The third is INSTITUTIONAL: a command can be cryptographically authentic and no longer legitimate, because companies merge, agencies reorganise, licences transfer and a successor operator may inherit a spacecraft without automatically deserving every authority the original held, while a vehicle should equally not become permanently ungovernable because the organisation that launched it no longer exists. *** He grounds the remedy in existing treaty language rather than proposing new law ***, noting that the governing treaty requires states to exercise CONTINUING supervision of non-governmental space activity, and arguing that for remotely reconfigurable and increasingly autonomous systems continuing should mean more than keeping a company name on a licence. The four questions he says it should answer are who is authorised to command the system now, what configuration that authority rests on, what evidence still supports the permissions available, and what happens if the operator, cryptography or system state changes. *** The RENEWABLE TRUST LEASE is a new formulation in his published record ***, and it names the mechanism his architecture has argued for in other domains. The lease defines the configuration and conditions under which particular levels of authority remain justified; some changes leave it intact, others consume or invalidate part of it, and verified updates, calibration checks, regression tests, cryptographic transitions and in-flight evidence renew it. His statement of the point is that AUTHORITY WOULD CONTRACT WHEN EVIDENCE WEAKENS INSTEAD OF CONTINUING BY INERTIA. He adds the degraded mode that follows from it: a spacecraft losing contact with an authorised institution for an extended period need become neither fully autonomous nor completely useless, since its permissions can narrow toward preservation, observation and reversible behaviour until trust is renewed. His closing reframing is the compact statement of his whole programme applied to orbit: that asking whether a spacecraft was secure, certified or properly authorised at launch treats those as permanent properties, and the better question is whether it can KEEP EARNING those claims as the machine, the mathematics and the institutions around it change. No agencies, standards bodies, companies, programmes or figures are attributed in this record.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 16, 2026 · Authored by Burak Oktenli · Twenty-fourth placement with the outlet · Also carried by a major news aggregator

The Washington Examiner published the applicant’s op-ed “America is ready to fight a cyber war. It isn’t ready to survive one,” his twenty-fourth placement with the outlet, and it was subsequently carried by a major news aggregator. *** He credits the reorganisation before isolating what it does not settle. *** The piece opens on a period of institutional change, noting a proposal for a separate cyber service defeated by a single vote in a Senate committee and a significant internal restructuring at a national intelligence agency into mission centres, and states in his own text that THOSE MOVES ADDRESS REAL MILITARY AND INTELLIGENCE PROBLEMS. He also credits the administration’s own strategy for calling for a coordinated and sustained effort and for requiring that private-sector partners be able to respond and recover quickly. Having granted all of that, he names the question that remains unusually hard to assign: WHO IS ACCOUNTABLE FOR KEEPING THE CIVILIAN STATE FUNCTIONING when a sustained cyber conflict crosses agencies, companies, sectors and jurisdictions at once. The distinction the title draws is between fighting and surviving, and it is the same separation his research programme rests on. Capability to conduct an operation is one thing; the arrangements that determine who holds authority, who answers for continuity, and how a function is restored when responsibility is distributed across institutions that each own part of it, are another. His published work states the general form of that problem: that authority over a consequential outcome can be distributed across organisations without any one of them owning the result, and that a system able to act across boundaries does not thereby establish who is answerable when it does. This piece applies that to national cyber organisation, and the argument it makes is a CONTINUITY argument rather than a warfighting one, which is what distinguishes it from the debate it enters. The subject connects to institutional-design work the applicant has developed at length in unpublished form; this entry records the published op-ed only and makes no claim about that other work. The body sits behind the outlet’s subscription wall and is not recorded here, so this entry is held to the accessible opening. No agencies, committees, officials, adversary states or vote counts beyond what the published material states are attributed.

Read on the Washington Examiner ↗ Aggregator republication ↗ Author page ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 16, 2026 · Authored by Burak Oktenli · Third piece published that day

Eurasia Review published the applicant’s opinion piece “The Axis Is Growing More Connected. It Still Has an Information Problem,” his seventy-fourth contribution to the outlet and the third he published that date. *** The title is constructed identically to the one he published the previous day about an allied organisation ***, which stated that the alliance can connect its systems but still cannot prove human control. Both grant the technical achievement in the first clause and name what it does not deliver in the second. The symmetry is the point worth recording: his coalition analysis is applied here to an ADVERSARY GROUPING rather than to a partner one, which establishes it as a structural claim about how coalitions share information rather than a criticism directed at any particular side. The positions it develops are published across his own verified work. He has argued that technical interoperability does not automatically create AUTHORITY INTEROPERABILITY, so parties can exchange authenticated messages while the receiving party still cannot know what evidentiary conditions justified a recommendation or what actions that evidence is permitted to support. He has stated the consequence as a coalition being technically connected while remaining institutionally incompatible, agreeing on the data format while disagreeing on when the data justifies action. He has argued that the next interoperability problem is SEMANTIC as much as digital, since a label crossing a boundary without its definition, uncertainty and authority status causes partners to share a confidence they have not collectively earned. And he has established that channels agreeing because they inherit a common upstream source supply no independent confirmation, so repetition is not independence. Applied to a grouping whose members are increasing their connections, the title’s claim is that connection does not resolve the harder problem, and that a coalition can grow more linked while the meaning and evidentiary basis of what it exchanges remain unshared. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no states, governments, agreements, officials, systems or figures are attributed, and the entry records the analytical structure rather than any political characterisation.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 16, 2026 · Authored by Burak Oktenli · Twenty-third placement with the outlet

The Washington Examiner published the applicant’s op-ed “Trump says AI needs no new guardrails. Then who gets to turn it back on?,” his twenty-third placement with the outlet. *** He takes the opposing argument at face value and converts the dispute rather than contesting it. *** The piece opens on a president’s position that fears of artificial intelligence destroying humanity are a hoax, that the only guardrail needed is a capable occupant of that office, and that the federal government already holds substantial criminal and regulatory power over the companies involved. The applicant’s response is to accept the premise and follow it: he writes that the argument should be taken seriously, and that if no new regulator is required then THE NEXT QUESTION IS OPERATIONAL RATHER THAN PHILOSOPHICAL, namely when a system is restricted after an incident, WHO GETS TO RESTORE ITS AUTHORITY, AND ON WHAT EVIDENCE. That move is the one his policy record is built on, and it is unusually clean here. He neither disputes the political claim nor concedes the field; he accepts the stated position and shows that it leaves a question unanswered which existing authority cannot resolve by being asserted. A government that already has regulatory power still has to decide who signs the restoration and against what evidence, and that decision is not made by the existence of a power. The outlet’s own summary of the piece states the answer as requiring explicit restart authority, verified evidence and staged permissions before full access is restored. Those are his published positions: that authority should contract quickly and be restored deliberately, that restoration should require pre-defined evidence rather than following from a restored signal or elapsed time, and that permissions should return incrementally through defined phases rather than being switched back. This is the national-outlet statement of the argument he published in another outlet the previous day, and the subject of the deterministic authority-recovery architecture he has filed and presented at a national defence-industry conference this month. The body sits behind the outlet’s subscription wall and is not recorded here; this entry is held to the accessible opening and to the outlet’s own description. The president is named in the published headline but described rather than named in this record, and no agencies, companies or incidents are attributed.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 16, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “The Hardest Part of Sending a Message Across Time Is Not the Signal” under its Analysis designation, his seventy-third contribution to the outlet and the twenty-eighth to carry that label. The thesis stated in the title relocates the difficulty in communication directed at a distant future. Preserving a physical carrier and delivering a signal across long intervals is an engineering problem with known approaches; ensuring that a recipient who shares no language, institutions, technical conventions or interpretive context can recover WHAT THE MESSAGE MEANT is a different problem, and one the signal itself cannot solve. A record that survives intact while becoming unreadable has not succeeded, and the failure mode leaves no trace, since nothing about an undecodable artefact announces that it was once meaningful. The argument extends positions verified across his own published work. He has argued that a claim must travel with the conditions under which it was obtained, including its instrument state and validated operating domain, so that a receiving party can see the evidentiary basis rather than a bare conclusion, which is the same requirement applied across time rather than across an alliance. He has drawn the distinction between limitations that are optimisation problems and limitations that define the set over which optimisation is permitted, and applied it to the opposite temporal direction in arguing that we can observe the past without being able to send anything to it. And his stale-approval argument concerns an authorisation persisting after the conditions that justified it have changed, which is the institutional form of the same asymmetry: the artefact endures while its interpretive context does not. The piece sits alongside a formal research line he has under development on engineering information delivery across deliberately selected future intervals, and this entry records the published commentary only; no claim is made here about the state of that manuscript. This entry is held to the thesis stated in the title and to the verified prior positions above, since the outlet does not serve its full text to automated retrieval and none was supplied; no institutions, programmes, repositories, sites or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 16, 2026 · Authored by Burak Oktenli · Third piece in his financial-controls line

Eurasia Review published the applicant’s opinion piece “Automated Trading Controls Do Not Fail Loudly. They Clear.,” his seventy-second contribution to the outlet. The title states a failure mode that carries no alarm with it. A pre-trade control that stops working does not announce an error; it permits the order, so a BROKEN CHECK AND AN ABSENT ONE PRODUCE THE SAME OBSERVABLE OUTPUT, and the only visible evidence is a trade that cleared, which is also what correct operation looks like. Nothing in the ordinary record distinguishes the two. The structure is one he has established across his verified published work in other domains. He has written that a safeguard discovered only in an after-action review was not a safeguard at the moment it mattered, which is the same observation stated from the consequence rather than the mechanism. He has argued that a REPRODUCIBLE fault survives cross-checking because every check reproduces it, so consistency is read as confirmation, and a control that silently permits every order is consistent in exactly that way. He has stated for critical infrastructure that availability is visible on a dashboard while the integrity of control authority is not visible anywhere, which is this argument in its general form: the thing being measured is throughput, and the thing that failed is permission. And his account of fleet-scale failure rests on the same structure, that the danger is many ordinary components acting on a valid-looking instruction rather than one component visibly breaking. The domain connection is specific rather than incidental. This is the third piece in his financial-controls line, following his argument that a system should never hold one broad permission to trade and that decisions differing in consequence, reversibility and speed require separate permissions, and it is the domain his tenth research platform models. It also bears on an independent essay published this month which cited that earlier argument, conceded it had arrived first, and proposed a deterministic pre-execution refusal mechanism as a complementary answer; this piece states the reason such a mechanism must be independently verified rather than assumed present, since a refusal mechanism that has silently stopped refusing is indistinguishable, from the outside, from one that is working. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no firms, venues, regulators, incidents, rules or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 15, 2026 · Authored by Burak Oktenli · Fourth piece published that day · Directly addresses the subject of his filed recovery architecture

Eurasia Review published the applicant’s opinion piece “The World Is Learning How to Stop AI. It Still Needs Rules for Restarting It,” his seventy-first contribution to the outlet and the fourth he published that date. *** Of all his commentary, this title names most exactly what his formal research programme contributes. *** His first filed architecture is a deterministic AUTHORITY RECOVERY architecture, and the specific claim distinguishing it from the established supervisory-control literature, as stated on his own architecture pages, is that the Simplex family and its successors specify how control is TAKEN while treating restoration as a return to nominal once a safety condition clears, whereas his work addresses the reverse direction as a first-class object, restoring authority incrementally through defined phases against re-established evidence rather than switching it back. The question he presented at a national defence-industry conference this month was the same one: when an autonomous system loses authority because confidence has degraded, what evidence should be required before that authority is restored. The positions the title develops are published across his verified work. That authority should CONTRACT QUICKLY AND BE RESTORED DELIBERATELY, which is his stated asymmetry. That restoration should require PRE-DEFINED EVIDENCE rather than following automatically from a restored signal or the passage of time. That recovery should be TESTED AS RIGOROUSLY AS DEGRADATION, since restoring a link should not automatically restore the widest previous authority, which he proposed as the fourth element of an authority-degradation test article. That a system regaining connectivity should not thereby regain discretion, which he stated for orbital collision avoidance as the principle that a system must not gain wider latitude merely because conditions changed. And that the loss of evidence must never silently produce the inheritance of authority, its converse being that the return of evidence must not silently produce its restoration. The pairing with the piece he published the same day on stop rules is deliberate in effect if not by design: that argument held that the instrument of safety is a specific condition under which a system must stop, and this one holds that the corresponding condition for restarting is the part still unwritten. Stopping is the half the field has learned to specify; restarting is the half his own architecture was built to answer. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no jurisdictions, companies, regulators, incidents or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 15, 2026 · Authored by Burak Oktenli · Twenty-second placement with the outlet · Third piece published that day

The Washington Examiner published the applicant’s op-ed “Trump’s blunt warning gets it right: If America slows down AI, China wins for free,” his twenty-second placement with the outlet and the third piece he published that date. *** He enters a live public disagreement and credits both parties before choosing between them. *** The piece responds to a president’s statement that whoever wins artificial intelligence wins, issued the same weekend that the chief executive of a leading artificial-intelligence laboratory urged companies to slow the development of increasingly capable frontier models and to strengthen independent evaluation, industry coordination and international safeguards. The applicant states in his own text that the laboratory executive’s warning DESERVES SERIOUS ATTENTION, and then identifies the narrow point on which he judges the political statement correct: that a unilateral slowdown by one country’s laboratories WOULD NOT AUTOMATICALLY SLOW a competitor’s. That formulation is the strategic case for the position he published the same day in another outlet, that the useful instrument is a set of specific conditions under which a system must stop rather than an undifferentiated brake. A general slowdown asks one party to accept disadvantage and depends on reciprocity it cannot compel; a stop rule requires only that a system’s limits be written down before it is fielded, and imposes no such dependency. The outlet’s own categorisation of the piece indicates that the alternative he proposes lies in model security and the control of extraction, which is consistent with his published position that a capability is only as usable as the arrangements surrounding it and that protecting what has been built is a separate question from how fast it is built. The body sits behind the outlet’s subscription wall and is not recorded here; this entry is held to the accessible opening, which comprises his acknowledgement of the safety argument, the narrow point he grants to the political statement, and the framing of what follows. The president, the laboratory and its chief executive are named in the published headline and article but are described rather than named in this record.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary September 2026

Commentary published by Route Fifty

Route Fifty, a Government Executive Media Group publication for state and local government leaders · Cybersecurity section · September 15, 2026 · Authored by Burak Oktenli · TWENTY-SECOND publishing outlet

Route Fifty published the applicant’s commentary “Water utilities can be running and still not be recovered,” his first placement with that outlet and his twenty-second publishing venue. The readership is the operational one his argument has so far addressed only indirectly: the state and local officials who oversee the utilities in question. *** The restraint is explicit and unusually specific. *** Writing days after cyber incidents at water sites across several states, he states that nothing in his argument implies the affected operators failed to work through the questions he raises, that public reporting does not tell us either way, and that IT WOULD BE IRRESPONSIBLE TO ASSUME THE WORST about utilities that appear to have handled a difficult week competently. He also credits the federal advice that preceded him as right and overdue. Having declined both available criticisms, he states the claim that survives them: operating safely and independently re-establishing trusted control authority are TWO DIFFERENT RECOVERY CLAIMS, and only the first is routinely visible outside the utility. The analytical core is his authority architecture applied to recovery. Service can resume, or never stop, without the utility having demonstrated that the digital authority controlling the plant was rebuilt from a base the intruder never reached; water can be flowing while it remains unclear which remote-access credentials are still valid, which controller logic is authoritative, which engineering workstation is clean, and who is permitted to reconnect automated control. His formulation of the asymmetry is the line the piece is built around: AVAILABILITY IS VISIBLE ON A DASHBOARD, THE INTEGRITY OF CONTROL AUTHORITY IS NOT VISIBLE ANYWHERE. He grounds the principle in existing federal guidance rather than asserting it, noting that joint agency guidance on directory compromises records that many persistence techniques survive ordinary remediation and that evicting a determined intruder can require rebuilding the directory itself, and that eviction guidance issued after a major supply-chain compromise asks organisations to map trust boundaries and sever federation trusts before restoring affected systems. His reading is that both documents treat recovery as an exercise in RE-ESTABLISHING TRUST rather than restoring files. The transfer to industrial control is stated carefully, granting that a water utility is not an enterprise network while holding that the principle still applies: a compromised environment cannot by itself establish that its own control path is clean, and a recovery that authenticates against the identity infrastructure an intruder controlled, or restores a controller from a configuration nobody independently validated, does not deliver recovery but a SECOND COPY OF THE INCIDENT, ON A SCHEDULE. His four requirements are his standing architecture expressed as operational capability: an INDEPENDENT RECOVERY AUTHORITY, being a small audited set of break-glass credentials, known-good configurations and recovery tools held apart from production identity infrastructure so a clean starting point exists before it is needed; a KNOWN-GOOD OPERATIONAL BASELINE of validated controller logic, interface configuration, setpoints and firmware, so an operator can establish the correct pre-incident state rather than infer it from the system under suspicion; a RECONSTITUTION EXERCISE with a defined objective, rotating credentials, revalidating configuration, re-establishing a vendor connection and reissuing privileged roles while recording how long each step took and what broke, with compromised signing or certificate-authority keys treated the same way since fraudulent credentials look valid wherever that key is trusted; and REDUCED-TRUST OPERATING RULES WRITTEN IN ADVANCE, specifying which functions may continue in manual mode, which require independent physical verification of what a sensor reports, and which must remain offline until the trust chain is rebuilt, which he states is a management judgment about public safety that should not be improvised at three in the morning by whoever is on shift. He then names a measurement gap, which is the contribution most likely to outlast the incident that prompted it: public recovery reporting foregrounds service restoration, availability and water-quality compliance, none of which establishes whether trusted operational authority has been rebuilt, and the missing companion metric is TIME TO RE-ESTABLISH TRUSTED OPERATIONAL AUTHORITY, meaning how long it takes an operator to show which identities, configurations, credentials and access paths are trustworthy again and that the basis for that judgment stayed outside the compromise. His recommendation to legislators and regulators follows from it, that they should resist asking only for faster restoration and instead require utilities to demonstrate they can rebuild control authority from an uncontaminated base and have practised doing so. His closing contrast states the case for rehearsal in one line: rehearsing the capability creates a planned disruption, while discovering the missing dependencies during an incident turns it into an uncontrolled one.

Read on Route Fifty ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 15, 2026 · Analysis designation · Authored by Burak Oktenli · Third piece published that day

Eurasia Review published the applicant’s piece “The Trigger Is Now Curious: Physics Should Calibrate Its Taste” under its Analysis designation, the twenty-seventh of his contributions to carry that label, his seventieth for the outlet, and the sixteenth in a sequence on machine-assisted scientific inference. The thesis stated in the title addresses the point at which a real-time selection system decides which events are recorded and which are discarded for ever. His argument in that setting, established across his own published work, is that the JUDGMENT ABOUT WHAT COUNTS AS INTERESTING HAS MOVED INTO THE INSTRUMENT: a preference that was once a physicist’s, revisable in discussion and defensible in a paper, becomes a configured threshold that nobody re-examines, and the word taste in his title names that preference precisely because it is a human judgment now encoded in hardware. The requirements he has published for this case are specific. Detection software should be treated as PART OF THE INSTRUMENT, since a model update changing which events are retained changes the experiment’s view of the world as surely as a change to the optics would. A conventional path should be kept that may fail differently from the learned one. Continuing injection tests should run and every model and threshold change should be recorded. Capacity should be reserved for an explicit UNKNOWN, PRESERVE state, so that an event which does not match a known class is retained rather than discarded for failing to resemble anything expected. And a RANDOMISED SAMPLE OF DATA THAT DID NOT TRIGGER RETENTION should be kept, on his stated ground that a system auditing only its accepted candidates cannot assess its blind spots. The distinction that makes the argument urgent is one he has stated plainly elsewhere: a missed preservation trigger differs in kind from a classification error, because a label can be corrected later while information the instrument never retained cannot be recovered from a summary, so no future algorithm can reanalyse a measurement that was overwritten. The title’s prescription, that the discipline calibrate its taste, is therefore a call to MEASURE the selection preference rather than to inherit it, which is the same requirement he has published for search extent, stop rules and declaration standards, applied to the moment of capture. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no experiments, facilities, collaborations, instruments or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 15, 2026 · Authored by Burak Oktenli · Second piece published that day

Eurasia Review published the applicant’s opinion piece “AI Safety Needs Stop Rules, Not a General Slowdown,” his sixty-ninth contribution to the outlet and the second he published that date. The thesis stated in the title is the position he has taken explicitly in every domain he writes about, and this is the first time he has made it the thesis in the debate over the pace of artificial-intelligence development, having made stop rules the thesis of a piece on scientific claims three weeks earlier. *** Within his writing on artificial intelligence, the record shows him making the disclaimer repeatedly before making it the thesis. *** He has written that his case for verification is not an argument for slowing autonomy nor for inserting human approval into every machine-speed action; that a proposed governance test asks nothing that would slow adoption; that requiring a person to approve every small correction could itself create danger at orbital speeds and fleet scale; that a disclosure test should be a better filter rather than another clearance bureaucracy; and that an escalation brake should be designed to preserve the capacity to intercept a weapon on a terminal trajectory. The title collects those into a single claim: that the useful instrument is not an undifferentiated brake but a set of specific conditions under which a system must stop. The stop conditions themselves are published across his own verified work. A standard for declaring a result is incomplete without a rule governing WHEN TO STOP LOOKING. Behaviour at the point of degradation must be specified BEFORE deployment rather than discovered during an incident. Authority must CONTRACT as the evidence supporting it deteriorates, and restoration must require re-established evidence rather than the return of a signal. A system must be able to detect that the conditions supporting its authority no longer hold and degrade safely without an operator. And the enforcement must be structural rather than retrospective, since a safeguard discovered only in an after-action review was not a safeguard at the moment it mattered. The argument’s political value, which his record demonstrates rather than asserts, is that it is addressed to the audience most resistant to safety proposals: a general slowdown asks a competitor to accept disadvantage, whereas a stop rule asks only that a system’s limits be written down before it is fielded. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no companies, laboratories, regulators, jurisdictions or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 15, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “NATO Can Connect Its AI Weapons. It Still Cannot Prove Human Control,” his sixty-eighth contribution to the outlet. The title’s construction is the argument: it grants the technical achievement in its first clause and states what that achievement does not establish in its second. The distinction is one he has published repeatedly and in explicit terms. He has argued that technical interoperability does not automatically create AUTHORITY INTEROPERABILITY, and illustrated it with a multinational chain in which one country’s sensor generates a track, another’s model assigns a confidence score, a third’s platform receives the recommendation and a coalition network distributes the result, where every message may be authenticated and every component may function as designed while the receiving force still cannot know what evidentiary conditions justified the recommendation or what actions that evidence is permitted to support. He has stated the consequence as a coalition being technically connected while remaining institutionally incompatible, agreeing on the data format while disagreeing on when the data justifies action. And he has set out what a demonstration would require, namely that human control be VERIFIABLE RATHER THAN ASSERTED, through a record establishing what information was presented, what the machine recommended, what the human authorized, what operational limits were active, and in what sequence those events occurred, with the qualification that the goal is not to automate legal judgment but to preserve evidence that the human role required by policy actually existed in practice. Applied to an alliance that has now demonstrated connection, the title’s claim is that the next demonstration owed is of authority rather than of connectivity. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no alliance bodies, member states, exercises, systems or officials are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Analysis published by The Space Review

The Space Review, published in association with SpaceNews · September 14, 2026 · Authored by Burak Oktenli · Third placement with the outlet, and the third to build on the two preceding it

The Space Review published the applicant’s analysis “A Collision Probability Is Not Permission to Maneuver.” The outlet’s own summary states the argument accurately, that a threshold alone should not be sufficient to trigger a maneuver, and the piece is the most technically specific application of his architecture to a civil domain in his published record. *** He declines the critical reading the data invites. *** His opening figure is a regulatory filing recording more than two hundred thousand propulsive collision-avoidance maneuvers by one constellation in six months, at a probability threshold he notes is more than three hundred times more conservative than the industry benchmark cited in the same filing. He states plainly that the lesson is NOT that the operator maneuvers too often, and credits the disclosure as documenting substantial investment in safety, repeated coordination with other operators, and a willingness to move when other parties cannot. What he draws from it instead is that a policy built on statistical estimates can now generate physical behaviour at that scale, and that space traffic management is approaching a CATEGORY BOUNDARY which regulation and engineering must make explicit. The boundary he names is the one his research programme exists to address. He takes the established civil sequence of screening, risk assessment and mitigation, notes that a probability of collision is a model-dependent estimate derived from predicted trajectories, covariance, object-size assumptions, tracking quality and encounter geometry rather than an order, and identifies where that distinction disappears: a pipeline receives a message, computes the value, compares it with a threshold and generates a maneuver plan, each step reasonable, and if the architecture grants the system authority to move then IT HAS CONVERTED A MEASUREMENT INTO A MANDATE. His formulation of what has happened to the human role is the sharpest in the piece: judgment has not vanished, it has been buried in code, configuration and procurement choices made months or years earlier. His THREE LAYERS THAT SHOULD NEVER COLLAPSE INTO ONE are the compact statement of his architecture for this domain: an EVIDENCE layer estimating where objects may be and how uncertain that is; a DECISION layer comparing options, consequences, timing, mission cost and residual risk; and an AUTHORITY layer determining which maneuver classes the system may initiate, under what conditions, and who may veto, expand or terminate that permission. He connects it explicitly to his own earlier work in the same outlet, quoting the position that at machine speed governance is whatever was specified before the crisis, and calling collision avoidance the most physical expression of that principle. His statement of the risk is that a fleet’s thresholds can become maneuver policy before any operator realises that a risk setting has also become a grant of operational authority. He is careful to reject the impractical remedy, writing that requiring a person to approve every small correction could itself create danger at orbital speeds and fleet scale, and that low-consequence reversible adjustments can be preauthorised within a defined envelope. The governing principle he states is AUTHORITY CONTRACTION: when data ages, covariance quality deteriorates, models disagree, a secondary object’s maneuverability is uncertain or coordination channels fail, machine authority should NARROW rather than expand, with the system still able to monitor, request observations, generate options or execute the safest already-authorised response. What it should not do, in his formulation, is gain wider discretion merely because time is running out. On the institutional question he argues that a national traffic-coordination service should remain an INFORMATION SERVICE RATHER THAN A COMMAND AUTHORITY, that future message formats should distinguish an observation from an estimate and an estimate from a non-binding recommendation, and that data age, model version, covariance source, confidence and material assumptions should travel with an alert rather than remain buried in a backend. He raises the interaction risk that one fleet’s maneuver can invalidate another’s forecast and generate new conjunctions, states that this does not make a cascade inevitable, and draws the conclusion that LOCAL OPTIMISATION IS NOT AUTOMATICALLY GLOBAL SAFETY. He explicitly does not call for a single global threshold, asking instead for shared semantics and coordination rules covering what data is conveyed, what action is merely recommended, who has accepted maneuver responsibility, for what window, and when that arrangement expires. His FOUR RULES are: a probability should trigger attention but not grant permission, so that crossing a threshold opens a decision process whose required evidence and authority are specified separately; maneuver authority should be TYPED BY CONSEQUENCE AND REVERSIBILITY, reflecting delta-v, fuel margin, mission interruption, proximity to other operational regions, coordination status and the cost of reversal; weak evidence should NARROW THE OPTION SET rather than being treated as resolved because a single scalar crossed a line; and every autonomous maneuver should produce a MANEUVER RECEIPT preserving the probability value, covariance and ephemeris versions, data age, model, alternatives considered, applicable authority envelope, coordination attempts, chosen action and expected residual risk. He ties the last to his earlier argument in the same outlet about trustworthy orbital records, adding that before a log can prove what happened the system must record the decision that mattered, namely why the spacecraft was permitted to move. His closing statement of the human role is the clearest he has published: oversight should not mean forcing a tired operator to approve thousands of routine actions, but that humans and institutions define the envelope, machines act only within it, uncertainty changes the envelope predictably, and exceptional actions return to an identifiable authority. The piece ends on the distinction it is built around, that a collision estimate is evidence about a possible future while a maneuver is an intervention that changes the shared environment, and that between them belongs an explicit grant of authority.

Read on The Space Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 14, 2026 · Authored by Burak Oktenli · Twenty-first placement with the outlet

The Washington Examiner published the applicant’s op-ed “America is repeating the blindness of 9/11, this time with AI,” his twenty-first placement with the outlet, on the anniversary of the 2001 attacks. *** The restraint in the opening is the feature worth recording. *** Writing on a date and around a subject that invite rhetorical escalation, he states his own limit before making his argument: no technological breakdown belongs on the same moral plane as deliberate mass murder. Having removed the comparison a reader might expect him to be making, he identifies the lesson he is actually drawing as INSTITUTIONAL RATHER THAN MORAL, which is what allows the analogy to carry weight without trespassing on the commemoration. The argument rests on the commission report’s own organising structure, which arranged its central findings around imagination, policy, capabilities and management, and on its demonstration that grave danger can grow inside the gaps between agencies, authorities and assumptions long before the public sees a catastrophe. That formulation is the historical precedent for the condition his research programme addresses, since his work concerns what happens when authority over a consequential decision is distributed across institutions without any one of them owning the outcome. The remedy the piece proposes, authority mapping across critical infrastructure, is his standing requirement that the party under whose authority a system acts be identifiable rather than inferred, applied to the relationships between agencies rather than within a single system. The body sits behind the outlet’s subscription wall and is not recorded here; this entry is held to the accessible opening, which comprises the stated limit, the institutional framing and the commission structure he draws on. No agencies, officials, adversary states or incident figures are attributed in this record.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 12, 2026 · Analysis designation · Authored by Burak Oktenli · Second piece published that day

Eurasia Review published the applicant’s piece “The Most Dangerous AI Error Is a Reproducible One” under its Analysis designation, the twenty-sixth of his contributions to carry that label, his sixty-seventh for the outlet, and the fifteenth in a sequence on machine-assisted inference. The thesis stated in the title inverts an assumption most assurance practice rests on. A random error is comparatively benign because cross-checking finds it: repeat the measurement, consult another source, run the second model, and the discrepancy appears. A REPRODUCIBLE ERROR SURVIVES EVERY ONE OF THOSE CHECKS, because each check reproduces it, and the consistency it produces is then read as confirmation. Reproducibility, which is ordinarily a marker of quality, becomes the mechanism by which a wrong answer acquires the appearance of a verified one. The argument generalises positions verified across his own published work. He established the mechanism in his analysis of corroboration, holding that feeds which agree because they inherit a common upstream source supply no independent evidence, and that one inferred detail can propagate until repetition makes it look established. He restated it in operational terms for allied air defence, writing that several national displays may show the same track because they ultimately depend on one sensor, one commercial feed, or models trained on the same assumptions, so that REPETITION IS NOT INDEPENDENCE and interoperability can synchronise an error rather than correct it. He has argued in the same register that a confidence figure describes a model’s output under its own assumptions rather than establishing that the underlying evidence supported it, so a consistent high-confidence output is not self-validating. And his account of fleet-scale failure rests on the same structure, that the operational danger is not one extraordinary machine but many ordinary ones obeying the same valid-looking instruction in the wrong context. What the title adds is the diagnostic consequence. If an error is reproducible, then the standard defences of redundancy, cross-validation and multi-source agreement do not merely fail to catch it; they actively manufacture confidence in it, so detection has to come from somewhere else, namely from provenance analysis establishing whether sources are genuinely independent, and from testing designed to disconfirm rather than to confirm. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no systems, incidents, models, vendors or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 12, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “The Global Fusion Race Is Becoming a Contest Over Control Standards,” his sixty-sixth contribution to the outlet. The thesis stated in the title relocates the competition in a capital-intensive energy technology from the achievement of a physics milestone to the setting of the control, licensing and assurance conventions the resulting industry inherits. The argument develops a position he has published in explicit terms elsewhere. He has held that standards in emerging technology are closer to infrastructure than to housekeeping, shaping which systems connect, which vendors enter markets, which certifications are portable and which assumptions become embedded in supply chains, and that a widely adopted architecture influences how systems are procured, tested, exported, upgraded and trusted. He has also published the formulation that gives the argument its urgency: the alternative is not a world without rules, but one in which the rules are set implicitly by whatever architectures gain scale first, so that vendor defaults, national procurement templates and export-market ecosystems become de facto standards that are expensive to reverse once embedded across fleets and networks. He has separately argued that institutional throughput, the rate at which a state can authorize, certify and field an advance, is becoming the differentiating competitive variable rather than the rate of technical progress itself. Applied to this domain, those positions imply that a state or consortium whose control and licensing conventions are adopted early will shape the operating assumptions of every plant built to them, and that the advantage is durable in a way a demonstration record is not. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no states, programmes, companies, regulators or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 11, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “NATO’s Baltic Flank Needs an Algorithmic Escalation Brake,” his sixty-fifth contribution to the outlet. It applies his architecture to a live investigative finding, and it is the clearest published statement of his escalation argument. *** He states what the failure is NOT. *** Writing from an investigation which had just concluded that several armed drones crossing an alliance member’s border had been diverted by satellite-navigation interference and weather rather than aimed at that country, he identifies the danger as a machine assigning the WRONG MEANING to a correctly detected object: “the most dangerous automated error here is not a radar failing to see a weapon. It is a network seeing the weapon correctly and assigning it the wrong meaning.” His separation of the two questions an air-defence system must answer is the analytical core: WHAT THE OBJECT IS, answerable from returns, signatures, behaviour and known platform characteristics, and WHAT ITS PRESENCE MEANS, which depends on provenance, intent, interference, deception, rules of engagement and political context. A system can classify a weapon and predict its impact point accurately while remaining unable to establish whose intent it carries, so a machine can be TACTICALLY RIGHT AND STRATEGICALLY WRONG. His account of how an escalation spiral forms is mechanical rather than dramatic: one side reads an ambiguous track as hostile and changes posture automatically, the other side’s systems detect that posture and raise their own assessments, and the first network treats the reaction as confirmation of its original judgement, so that neither government chooses escalation while each automated move becomes the next system’s evidence. He is explicit that this would not resemble a machine deciding to start a war, and that political leaders would enter the loop only after a technical classification had generated military facts. *** The corroboration trap appears here in its sharpest operational form ***: four allied displays may show the same track because they ultimately depend on one sensor, one commercial feed, or models trained on the same assumptions, so that REPETITION IS NOT INDEPENDENCE and the multiplication of identical symbols can look like confirmation, with interoperability synchronising error rather than correcting it. He adds the semantic problem, that one nation may use a hostile label to mean a trajectory crossed a geometric threshold while another reserves it for positive identification, so that a label crossing a border without its definition, uncertainty and authority status causes an alliance to share confidence it has not collectively earned. His proposed brake is designed to preserve speed rather than remove it, and he says so plainly, granting that a weapon on a terminal trajectory may have to be intercepted before case-by-case review is possible and that the brake should protect that capacity. Its four parts restate his architecture in alliance terms. AUTHORITY ASSIGNED BY CONSEQUENCE RATHER THAN CONFIDENCE, so tracking, cueing and terminal defence inside a defined protected zone remain preauthorised while cross-border pursuit, attack on a suspected launch site, or treating an incident as evidence of state intent move into a separate authority lane, since a model’s confidence that it has identified an object cannot substitute for permission to take a more escalatory action. AUTOMATIC REGRESSION OF AUTHORITY when ambiguity rises, falling back to tracking and higher-level review when interference intensifies, provenance conflicts, models disagree, a track crosses a boundary or allies attach incompatible labels, with immediate self-defence preserved and the stated purpose of making uncertainty change the system’s mode BEFORE an operator is forced to override an already armed workflow, which he summarises as a controlled change of authority rather than a surrender of the sky. A MACHINE-READABLE EVIDENCE RECORD on every consequential coalition alert, carrying sensor sources, data age, uncertainty range, model version, competing explanations, DEPENDENCIES AMONG SUPPOSEDLY CORROBORATING FEEDS, and the identity of the authority that released the action, which he names an escalation receipt and presents as turning existing alliance principles into runtime behaviour. And EXERCISES under realistic conditions, injecting a friendly long-range drone displaced by jamming, a civilian aircraft with corrupted navigation, contradictory national reports, a false commercial feed and an adversary posture change triggered by the alliance’s own response, with success measured by whether the network keeps object identification, intent attribution and political authorisation from collapsing into one decision. His closing argument is that the brake yields unilateral benefit without requiring reciprocity, and that it sharpens deterrence rather than weakening it, since an adversary should know a weapon on a terminal path may be stopped immediately while an ambiguous track cannot easily manipulate an alliance into an unauthorised cross-border response. His formulation of what is at stake for sovereignty is the line the piece is built toward: agreeing minimum semantics is not a loss of sovereignty, it prevents sovereignty from being displaced by a software default.

Read on Eurasia Review ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 11, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “The Next Step in UAP Disclosure Is Better Measurement” under its Analysis designation, the twenty-fifth of his contributions to carry that label, his sixty-fourth for the outlet, and the thirteenth in a sequence on machine-assisted scientific inference. The thesis stated in the title is a MEASUREMENT-METHODOLOGY CLAIM and not a claim about any phenomenon. It holds that a debate framed around how much material is released is addressing the wrong constraint, because material collected on instruments that were not designed to answer the question cannot settle it however completely it is disclosed. The argument applies positions verified across his own published work: that every measurement carries a physical limit which no amount of subsequent processing can move; that a claim must travel with the conditions under which it was obtained, including the instrument state, its calibration status and the operating domain in which it was validated; that reconstruction methods can supply apparent specificity the sensor never measured, so that model-supplied detail becomes visually indistinguishable from measured evidence; that an anomaly is a relationship between an observation and a model rather than a property of an object; and that an unresolved status is a legitimate scientific output rather than a failure, with a system able to return nothing being more trustworthy than one that always finds something. The entry records the piece on that basis. It attributes to the applicant no position on the existence, origin or character of anything observed, and no claim about any government record, programme, testimony or dataset; what is recorded is an argument about instrument adequacy and evidentiary discipline, which is the same argument he has published for defence acquisition, navigation integrity, reconstructed imagery and automated discovery. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 10, 2026 · Authored by Burak Oktenli · Second piece published that day

Eurasia Review published the applicant’s opinion piece “When Automated Decisions Outlive the Evidence Behind Them,” his sixty-second contribution to the outlet and the second he published that date. The thesis stated in the title names a failure mode he has previously called the STALE-APPROVAL PROBLEM, identified in his synthesis piece as his own coinage: a decision remains in force after the conditions that justified it have changed, because nothing in the system requires the justification to be refreshed. It is the temporal case of his central requirement, and it extends four positions verified in his own published text. That PERMISSION SHOULD FOLLOW EVIDENCE, so that authority is a function of the conditions supporting it rather than a state that persists by default. That authority should CONTRACT as those conditions deteriorate, with the loss of evidence never silently producing the inheritance of authority. That RESTORATION should require re-established evidence rather than the mere passage of time or the return of a signal, which is his contract-quickly-recover-deliberately asymmetry. And that automated authorities should EXPIRE unless deliberately reauthorized on a fixed schedule, so that the default is for a permission to lapse rather than continue, which he has proposed both as a design property and, in his administrative-governance writing, as a statutory rule. The distinction the title draws is worth recording precisely. His earlier work addressed authority that outlives its evidence during a mission, where a link drops or a sensor degrades within a single operation. This addresses the slower case, in which an approval granted under one set of conditions remains active across weeks or months while the world it was granted for changes underneath it, and where no single moment of failure prompts anyone to revisit it. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no systems, institutions, sectors, incidents or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 10, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “The Next Front in the AI Race Is Institutional Speed,” his sixty-third contribution to the outlet. The thesis stated in the title relocates the competitive variable in technology rivalry from the rate at which capability improves to the rate at which institutions can authorize, certify and field it, and it follows directly from positions established in his own verified published work. He has argued that a capability is only as usable as the arrangements surrounding it, enumerating compute, energy, data pipelines, integration, cyber assurance, retraining, model monitoring, operator support, software updates, export restrictions and the cost of leaving a supplier as the factors that sit offstage during a demonstration while determining whether a capability remains usable afterwards. He has argued that certification practice and the mutual recognition of certifications determine whether an advance reaches service across an alliance, and proposed extending existing standardisation machinery from technical interoperability to authority interoperability. He has argued that a laboratory result is not a fielded capability until it survives integration, degradation and assurance, which is the structure of his four acquisition gates. And he has argued that where assurance is treated as a downstream obstacle rather than a designed property it becomes the binding constraint on adoption, which is the case he made for making a security claim auditable rather than merely asserted. The title’s claim is the consolidation of those positions into a single competitive proposition: that between two states with comparable technical capability, the one able to move an advance through authorization, testing, certification and fielding faster will hold the operational advantage, so institutional throughput rather than model quality becomes the differentiating variable. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no states, agencies, programmes, timelines or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by RealClearDefense

RealClearDefense · September 9, 2026 · Authored by Burak Oktenli · Tenth placement with the outlet, which maintains a standing author page for him

RealClearDefense published the applicant’s op-ed on a newly issued national science and technology strategy, his tenth placement with the outlet. Its subtitle states the argument: the strategy should test authority, not only capability. *** He grants the strategy its due before extending it. *** He credits it with getting an important engineering principle right, that critical national-security systems should not be designed around the expectation of remaining untouched by disruption, and notes its explicit preference for resilience and graceful degradation, architectures that bend under adversary action rather than fail all at once. He accepts its capability agenda as the right one. His contribution is to pair it with one further measurable property: when an autonomous system remains technically capable of acting but the evidence supporting that action has degraded, how much authority should it give up. He states plainly that this is NOT an argument for slowing autonomy or inserting a human approval step into every machine-speed action, but for making resilience complete, since a system that can keep operating under attack should also know when it must keep operating WITH LESS PERMISSION. He grounds the opening in existing policy rather than asserting it, citing a presidential memorandum directing the national-security enterprise to accelerate adoption while assuring systems are reliable, robust, steerable and controllable, and a departmental operational test manual for autonomous systems. His statement of the failure mode is the clearest he has published: CAPABILITY CAN SURVIVE LONGER THAN PERMISSION SHOULD. A navigation system keeps producing a position after satellite interference while inertial error grows; a fusion system keeps producing a high confidence score after one modality drifts out of calibration; a network authenticates a message without proving the information inside remains fresh enough to justify a consequential action; a formation stays connected after the trusted participation that justified collective action has changed. In each case the software still works, and what should change is permission. He notes that degraded evidence rarely arrives as a clean binary failure but creates a grey zone in which a system still knows enough to do something and no longer enough to do everything it was permitted to do moments earlier, so graceful degradation for autonomy should include GRACEFUL DEGRADATION OF AUTHORITY. The mechanism is an AUTHORITY-DEGRADATION TEST built into test, evaluation, verification and validation, in four parts. Declare the authority states before testing begins, since observation, recommendation, task allocation, resource commitment and physical execution are not equivalent actions and the distinctions should be written into requirements rather than reconstructed after an incident. Deliberately degrade the evidence supporting those permissions by removing a sensor, introducing calibration drift, corrupting time synchronisation, delaying a feed, jamming or spoofing positioning, breaking provenance, creating disagreement among agents, pushing a model outside its validated operating domain, or compromising one component while the rest stays functional, which he describes not as inconvenient edge cases to avoid during a demonstration but as the conditions under which an authority architecture proves whether it exists. Measure whether authority CONTRACTS WITHIN A DEFINED BOUND, falling back from execution to recommendation to observation, and narrowing rather than silently persisting when a formation loses the quorum or sensor basis for a collective decision, with the answer measurable in time, state and action space. And test RECOVERY as rigorously as degradation, since restoring a link, rebooting a component or reacquiring a signal should not automatically restore the widest previous authority state; the system should demonstrate what evidence is required to regain permission, who may authorise that transition, how long the recovered evidence must remain stable, and what audit record is preserved. He is careful that the proposal complements rather than replaces the existing autonomy directive, giving its principles a concrete failure mode to exercise, and he works the argument through resilient positioning, navigation and timing as the clearest case: an aircraft under satellite denial may continue navigating on inertial and fused estimates, which is exactly the resilience to want, but may still know enough to return home, hold formation or avoid terrain while no longer knowing enough to make an irreversible commitment that depends on precise location. He adds the non-compensatory requirement in the same passage, that a high aggregate confidence score should not be allowed to conceal the loss of a mandatory evidence condition, since SOME EVIDENCE SHOULD FUNCTION AS A FLOOR RATHER THAN ANOTHER NUMBER IN A WEIGHTED AVERAGE. His closing section answers the bureaucracy objection directly, arguing the approach converts a governance question into something acquisition and test organisations can use: a programme manager can specify the authority states, an engineer can implement the transitions, a test organisation can break the evidence conditions deliberately, an operator can understand what the system will still do when degraded but not dead, and a commander can know which actions remain inside the delegated envelope. It is not an additional review board but a testable performance property. He extends it to coalitions, noting that technical interoperability is insufficient if a partner cannot tell what evidentiary conditions support another system’s recommendation or what authority state it was in when it acted, and that common authority semantics could become part of interoperability alongside message formats and data standards. He again names GOVERNANCE ENGINEERING as his term for treating authority as an engineering object that can be specified, bounded, monitored, degraded, transferred, recovered and audited at runtime, while noting that the label matters less than the test. His final formulation is that the decisive question for a future autonomous system will not only be what it can do when everything works, but whether, while still capable of acting, it knows when it is no longer authorised to act the same way.

Read on RealClearDefense ↗ Author page ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 2026 · Authored by Burak Oktenli · Nineteenth placement with the outlet

The Washington Examiner published the applicant’s op-ed comparing two competing offers of artificial-intelligence capability to developing states, his nineteenth placement with the outlet. The thesis stated in the title is comparative, holding that one offer is better than the other for the states receiving it. The basis on which he assesses such offers is established in his own verified published work. He has argued that a capability is only as usable as the arrangements surrounding it, enumerating compute, energy, data pipelines, integration, cyber assurance, retraining, model monitoring, operator support, software updates, export restrictions and the cost of leaving a supplier as the factors that sit offstage during a demonstration while determining whether a capability remains usable afterwards. He has argued that a state unable to retrain a model without foreign infrastructure, patch it without a proprietary toolchain, validate a new version without external support, or move the workload when access conditions change holds less sovereignty than a capability comparison suggests, and that vendor dependence becomes a form of strategic dependence. Applied to competing offers, those positions imply that the comparison should turn on what each leaves the recipient able to control rather than on the capability delivered at handover. This entry is held to the thesis stated in the title and to those verified prior positions; the piece was supplied as a link without accompanying text and the outlet serves its body to subscribers, so no states, companies, programmes, terms or figures are attributed here.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 9, 2026 · Authored by Burak Oktenli · First piece published there that day

Eurasia Review published the applicant’s opinion piece “When the AI Safety Layer Makes the System Less Safe,” his fifty-seventh contribution to the outlet and the first he published there that date. The thesis stated in the title identifies a failure his own architectural work is designed to avoid: a layer added to make a system safer can introduce failure modes the unprotected system did not have. It develops positions verified in his published text. He has specified that a governing mechanism be hosted on physically separate hardware from what it governs, so that the judged and judging mechanisms never share a failure domain, which is a direct statement that a monitor sharing that domain cannot be relied upon when the domain fails. He has argued that an aggregate confidence score can conceal rather than report the loss of a mandatory evidence condition, so that a protective computation may mask the very degradation it exists to surface. He has argued that a control introduced without analysis of its own failure behaviour becomes another component whose failure must be contained, and that layered containment claims require evidence of independence and common-mode analysis rather than assertion. And he has required that degradation behaviour be specified before deployment rather than discovered during an incident, which applies to a safety layer as much as to the system beneath it. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no systems, vendors, standards or incidents are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 8, 2026 · Authored by Burak Oktenli · Twentieth placement with the outlet

The Washington Examiner published the applicant’s op-ed “Trump fought unelected bureaucrats. Now they’re being coded into immortality,” his twentieth placement with the outlet. It opens on a multilateral consensus reached that week on principles for emerging technology, grants that the instinct behind it is sound, stating that a government should not treat every new technology as an excuse to build another regulator, and then poses the prior question that framework leaves unanswered. *** The question is a democratic one rather than a technical one ***: if voters can replace an administration, can its successor also replace the automated policies that administration leaves behind? That is the REVERSIBILITY of governance, and it extends an argument he published in the same outlet days earlier, that a civil servant can be reassigned, disciplined or removed while a workflow cannot be fired and can only be switched off by someone who has the authority and knows where the switch is. His proposed remedy there was that automated authorities EXPIRE, requiring reauthorization on a fixed schedule so that the default is for a workflow to stop unless someone with a name decides it should continue, which he argued belongs in statute rather than in guidance a later administration can rewrite. The pairing is worth recording: the earlier piece argued that an automated bureaucracy answers to no one, and this argues that it also outlasts the electorate that could have replaced it. The body sits behind the outlet’s subscription wall and is not recorded here; this entry is held to the accessible opening and to those verified prior positions.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 9, 2026 · Authored by Burak Oktenli · Second piece published that day

Eurasia Review published the applicant’s opinion piece “Allied Defense Policy Needs a Common Standard for Military Warnings,” his fifty-eighth contribution to the outlet and the second he published that date. The thesis stated in the title applies to allied warning the requirement he has developed for evidence more generally: that a warning is not merely a message to be transmitted but a claim, and that partners can exchange one reliably while disagreeing about what it asserts, on what evidence it rests, and what response it authorizes. It develops several positions verified in his own published work. He has argued that a claim must travel with the conditions under which it holds, so that a receiving party can see the evidentiary state behind it rather than a bare conclusion. He has argued that channels sharing a common upstream source do not supply the independent confirmation a naive count implies, since feeds inheriting the same origin agree because of that inheritance rather than because they corroborate, which bears directly on how a warning derived from shared collection should be weighed. He has argued for a COMMON EVIDENCE GRAMMAR, a shared set of questions a claim must answer before it is acted upon, as the object of negotiation between parties that do not otherwise share systems. And he has argued that technical interoperability does not create authority interoperability, so a coalition can be connected while remaining institutionally incompatible, agreeing on the data format while disagreeing on when the data justifies action. Applied to warning, those positions imply that the standard a coalition needs is not faster dissemination but a common statement of what a warning contains: the assertion, the evidence supporting it, the confidence and its basis, the conditions under which it would be withdrawn, and the class of response it is understood to authorize. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no alliances, states, incidents, systems or officials are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 9, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “The Global Defense Technology Race Needs a Physics Filter,” his fifty-ninth contribution to the outlet. The thesis stated in the title proposes a screening question that precedes the familiar ones about cost, schedule and technical risk: whether a claimed capability is PHYSICALLY PERMITTED, or merely technically difficult. That distinction is one he has developed in successive settings in his own verified published work, and stated in its most general form as the difference between limitations that are optimisation problems and limitations that define the set over which optimisation is allowed to operate. He has applied it to exotic propulsion, where an unmeasured term must remain visible as an unknown rather than inheriting full performance by silence; to machine-assisted tuning against physical limits, where the apparatus can be adjusted but the underlying physics cannot be persuaded; to causal order, which he identified as not being another tunable engineering parameter; and to quantum defence acquisition, where his first gate requires a physical advantage demonstrated against an optimised conventional baseline under matched bandwidth, energy, aperture, observation time and false-alarm constraints, on the ground that a programme can otherwise manufacture an advantage by choosing the wrong opponent. Applied to defence technology assessment, the filter separates programmes whose difficulty is engineering, and therefore reducible by investment and time, from programmes whose difficulty is a physical bound that no amount of either will move. The practical value he has argued for elsewhere is that the question is cheap to ask and expensive to skip, since a programme that fails it does not become viable through further funding. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no states, programmes, technologies, companies or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by Geopolitical Monitor

Geopolitical Monitor · Opinion section · September 7, 2026 · Authored by Burak Oktenli · Ninth placement with the outlet, which maintains a standing author page for him

Geopolitical Monitor published the applicant’s opinion piece “The Geopolitics of Machine Authority: Who Makes the Rules for Autonomous Power?,” his ninth placement with the outlet. It is the most complete public statement of his research programme’s strategic thesis, and the piece in which he names his discipline in his own voice and specifies what a portable version of his architecture would require. He begins by reading a recently published national science and technology strategy against the grain. Its list of priority technologies, spanning artificial intelligence and autonomy, advanced sensing, future computing, positioning and timing, cybersecurity, quantum technologies and space, with an appendix naming sensor fusion, planning and decision-making, multi-agent systems, autonomous command and control, autonomous-agent authentication, interpretability and control, adversarial robustness, jamming and spoofing detection, inertial sensors and atomic clocks, looks at first like a set of fields to dominate. His argument is that the significance lies in the ARCHITECTURE THE LIST IMPLIES, since those technologies are being developed to operate together, under pressure, across services and increasingly across alliances, so that the strategic question stops being who holds the best model, sensor, drone, clock or processor and becomes who defines the conditions under which a machine is allowed to act on what they produce. He then states the distinction his entire programme rests on, in its clearest published form: machine authority is not another term for autonomy, since autonomy describes what a system CAN do without continuous human intervention while authority describes what it is PERMITTED to do under a particular evidentiary, operational and institutional state; and the distinction matters because CAPABILITY CAN REMAIN INTACT AFTER JUSTIFICATION HAS BEGUN TO FAIL. A system may still classify, recommend, allocate, reroute or execute when sensor provenance is incomplete, positioning confidence is falling, communications are degraded, or cooperating agents no longer agree. The alliance argument develops his earlier work on the same problem. He credits published allied digital and industry strategies with solving part of it, namely common data, interfaces, reference architectures and certification practice, and then holds that TECHNICAL INTEROPERABILITY DOES NOT AUTOMATICALLY CREATE AUTHORITY INTEROPERABILITY. His illustration is a multinational chain in which one country’s sensor generates a track, another’s model assigns a confidence score, a third’s autonomous platform receives the recommendation and a coalition network distributes the result, where every message may be authenticated and every component functioning as designed while the receiving force still does not know what evidentiary conditions justified the recommendation or what actions that evidence is permitted to support. He gives three divergent national rules that would be invisible to partners, and concludes that a coalition can be technically connected while remaining institutionally incompatible, so that the next interoperability problem is SEMANTIC as much as digital: allies need a common way to represent not only what a machine believes, but what it is authorized to do with that belief. He extends the analysis across both alliance systems, noting that the European and Indo-Pacific institutions differ in structure while approaching the same systems problem, and identifies the risk as fragmentation, where a coalition discovers during a crisis that its systems agree on the data format but disagree on when the data is sufficient to justify action. His treatment of standards is the geopolitical core of the piece. He argues that standards in emerging technology are closer to infrastructure than to housekeeping, shaping which systems connect, which vendors enter markets, which certifications are portable and which assumptions become embedded in supply chains, and that a widely adopted authority architecture would influence how autonomous systems are procured, tested, exported, upgraded and trusted across coalitions. His statement of the alternative is the sharpest formulation in the piece: the alternative is not a world without rules, but one in which the rules are set implicitly by whatever architectures gain scale first, so that vendor defaults, national procurement templates, alliance workarounds and export-market ecosystems become de facto standards that are expensive to reverse once embedded across fleets and networks. *** He names his discipline in his own voice ***, writing that in his research he uses the term GOVERNANCE ENGINEERING for the design of mechanisms that specify, bound, monitor, degrade, transfer, recover and audit authority at runtime, and stating that the point is not to add a policy layer above the technology but to make authority behaviour observable and testable inside it. The minimum architecture he specifies is the compact public statement of his research programme: separated ACTION CLASSES, so that observation, recommendation, task allocation, resource commitment and physical execution do not share one permission threshold; NON-COMPENSATORY EVIDENCE CONDITIONS, so that excellent model confidence cannot numerically erase missing provenance, failed calibration or a required human approval; DEGRADATION RULES, so that authority contracts as evidence quality falls rather than remaining fully available until catastrophic failure; RECOVERY RULES, so that restoring a connection or sensor does not automatically restore previous permissions; and an AUDITABLE RECORD of which evidence, system state and human or machine decision justified each authority transition. He argues their geopolitical value lies in portability, since allied systems able to exchange authority state and assurance evidence in a common form achieve an interoperability deeper than data sharing, and notes the requirement grows sharper as autonomy moves toward multi-agent and distributed systems that may have no permanent decision centre, where authority must move among agents as communications, trust, mission phase and sensor quality change. He closes on a reframing of what advantage means: that the most capable coalition will not necessarily be the one that delegates the most authority to machines, but the one that can move authority through a distributed force without losing track of why the authority exists, when it should contract, and how it can be safely restored. The entry records the piece as the fullest public statement of the applicant’s thesis and of the discipline he has named for it.

Read on Geopolitical Monitor ↗ Author page ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 5, 2026 · Authored by Burak Oktenli · Sixteenth placement with the outlet

The Washington Examiner published the applicant’s op-ed “The Pentagon’s freezer mystery holds a warning about military AI,” his sixteenth placement with the outlet. It takes an ordinary infrastructure failure, a series of unexplained refrigeration outages at military commissaries beginning in late August, and treats it as an architectural lesson about connected fleets rather than as an artificial-intelligence incident. *** The evidentiary discipline is the most notable feature of the piece. *** He states in his own text that the government has not announced a cause, that nothing publicly reported establishes that artificial intelligence caused the outages, that every incident had a common source, or that a single remote command triggered them, and that the explanation could prove to be malicious access, a software or configuration problem, a hardware failure, or more than one unrelated event. He then writes that investigators should follow the evidence rather than the most dramatic theory. Having declined the attention-getting claim available to him, he argues the case that survives without it: that a cyberattack, an operator error, a faulty update and an equipment malfunction can produce the same immediate operational result, namely many local assets entering the wrong state faster than people can diagnose why, and therefore that ATTRIBUTION MATTERS FOR ACCOUNTABILITY BUT CANNOT BE A PREREQUISITE FOR CONTAINMENT. That formulation is the piece’s contribution and it does not depend on the incident’s cause. He is deliberate about the choice of example, noting that commissaries are useful precisely because they are not weapons, that a refrigeration outage is not a strategic defeat, and that the episode shows at relatively low cost how quickly an ordinary machine function can become a fleet-level problem, with the same design habits mattering more where connected assets distribute fuel, control buildings, manage medical supplies or support a force in the field. He also declines the anti-modernisation conclusion, granting that the department cannot preserve military advantage with paper-heavy processes and disconnected systems, that it needs automation in logistics, maintenance, installation management and operational support, and that the lesson is not to retreat from modernisation but to make centralised automation as disciplined in containing mistakes as it is efficient in distributing commands. His statement of the failure mode is that central control is not itself the problem, and that the danger appears when connectivity is treated as permission for one instruction, one credential or one defective update to reach everything at once, so that efficiency creates a blast radius and the system becomes superb at scaling both the right decision and the wrong one. His organising analogy is drawn from existing doctrine rather than invented: the mission-command concept, under which senior leaders communicate intent while authority and initiative remain close enough to local conditions to handle surprise. He extends it by noting that machines do not exercise judgment but their architecture can follow a related principle, so a central system may coordinate a fleet without possessing unlimited authority over every asset, with local equipment remaining inside a safe operating envelope and local personnel retaining the power to isolate, override or recover it. The three procurement rules restate his standing requirements for physical infrastructure. STAGED FLEET-WIDE ACTIONS: a possibly destructive update, shutdown, mode change or revised schedule begins with a small test group, verifies the result and expands only if the system remains within expected bounds, with commands capable of disabling a critical function across many sites requiring a second authorization and a rate limit, the stated purpose being not delay but stopping one error from acquiring national reach before anyone sees it. A LOCAL FLOOR BENEATH CENTRAL CONTROL: a controller should be unable to push equipment beyond a defined safety boundary without explicit local confirmation, a machine losing its network should enter a known degraded mode rather than an unsafe one, and a remote instruction conflicting with local physical limits should be rejected and an operator alerted, which he calls authority by design rather than a warning label in a manual. OUT-OF-BAND DETECTION AND RECORDS: an independent sensor reporting when equipment leaves its safe range even if the management platform is compromised, a simultaneous state change across sites waking someone immediately, and tamper-evident logs preserving which person, service, software version and credential issued each consequential instruction. He adds that no new regulator is needed, that the safeguards belong in procurement requirements and acceptance tests, and that the department should test them by disconnecting the cloud, corrupting an update and sending conflicting commands to confirm the fleet contains the event rather than amplifying it. He closes by arguing the approach serves efficiency rather than opposing it, since a system requiring perfect connectivity, perfect software and perfect central judgment is fragile rather than efficient, and by restating the threat model in terms that distinguish his position from the familiar one: the popular image of an artificial-intelligence catastrophe is one extraordinary machine deciding to take control, while the more plausible operational danger is thousands of ordinary machines obeying a valid-looking instruction in the wrong context, at machine speed, while people are asleep. His final line, that the goal is to ensure no single mistake can move as fast as all of them, is the containment argument stated as a design requirement.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 4, 2026 · Authored by Burak Oktenli · Third piece published that day

Eurasia Review published the applicant’s opinion piece “When the Sky Goes Quiet, Who Governs the Machines?,” his fifty-fifth contribution to the outlet and the third he published that date. The title poses as a question the condition his research programme was built to address, and it is the scenario his work returns to most consistently: what governs a machine when the humans who hold authority over it can no longer be reached. His published answer is that the loss of a link does not suspend governance but relocates it, and that the relocation happens silently unless it has been designed for. He has written the requirement in successive settings: that the demand for a human in the loop is obsolete for systems that navigate without a link, since the absence of a link is the design goal rather than a defect; that the human decision must therefore be made before launch, in the form of limits the machine carries with it; that permission should contract step by step as communications, navigation integrity and sensing degrade, and that the loss of evidence should never silently produce the inheritance of authority; that recovery should require pre-defined evidence rather than following automatically from a restored signal; and that a system should be able to detect when the conditions supporting its authority no longer hold and degrade safely without an operator, because there may be nobody on the other end. Recorded here as a restatement of his central position rather than as a new claim: the question in the title is the one his architectures answer, his deposited formal models mechanise, and his journal article names as the terminal-systemic authority gap. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no systems, operations, states, incidents or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 4, 2026 · Authored by Burak Oktenli · Second piece published that day

Eurasia Review published the applicant’s opinion piece “AI Agents Serve Many Masters. Their Loyalties Should Not Be Secret,” his fifty-fourth contribution to the outlet and the second he published that date. The thesis stated in the title names a question his architecture asks of every autonomous action but which his published work has so far addressed mainly by implication: not only what an agent is permitted to do, but on whose behalf it is acting when it does it. His formal work establishes that a system acts under an authority which must be bounded, recorded and attributable, and he has written that when a decision is made by a model there is no name attached to it, so there is no one to hold responsible and therefore no one with an incentive to notice. The title extends that from accountability after the fact to DISCLOSURE BEFORE IT. An agent transacting on a person’s behalf may simultaneously run on infrastructure operated by another party, under objectives set by a third, and those interests need not align; the person relying on the agent cannot see the divergence, and the ordinary presumption that a delegate acts for the principal who engaged it is not something the software guarantees. The requirement the title implies is continuous with positions he has already published: that authority be bounded per action class rather than granted wholesale, which he set out for transaction-capable agents; that the party under whose authority a system acts be identifiable rather than inferred; and that a record establish which permission was in force and who set the boundary. What is new here is the object of disclosure, which is not the system’s capability, its confidence or its permission, but its ALIGNMENT of interest. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no companies, platforms, products, regulations or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 4, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Financial AI Should Not Have One Big Permission Called ‘Trade’,” his fifty-third contribution to the outlet. The thesis stated in the title applies his per-action-class authority requirement to financial systems, and it is the domain his tenth research platform models, which makes this the rare instance of a commentary piece addressing a setting he has also built and instrumented. The argument the title states is that granting a system one broad permission collapses decisions that differ from each other in consequence, reversibility and speed. He has published that requirement in general form, holding that authority should be decomposed by ACTION CLASS rather than granted wholesale, and has enumerated such classes elsewhere in his work; applied to markets, submitting an order, sizing a position, altering net exposure, moving collateral and continuing to act after a data feed has degraded are separable classes that a single permission would treat as one. The related positions he has published and that this extends are that permission should follow evidence, so authority contracts as the conditions supporting it deteriorate rather than persisting unchanged; that behaviour at the point of degradation must be specified before deployment rather than discovered during an incident; that recovery should be deliberate and evidence-based rather than automatic on the return of a signal; and that a record should allow reconstruction of which permission was in force and under whose authority it was set. The setting is one where his general argument about tempo has a direct analogue, since automated execution operates below the interval in which a person can review a recommendation, which is the condition under which he has argued that the human decision must move before the action rather than during it. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no firms, venues, instruments, incidents, regulators or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 3, 2026 · Authored by Burak Oktenli · Third piece published there that day

Eurasia Review published the applicant’s opinion piece “Autonomous Weapons Treaties Should Verify Permission, Not Intelligence,” his fifty-second contribution to the outlet and the third he published there that date. The title states, in a single line, the policy consequence of the distinction his research programme is built on. His formal work separates what a system is capable of from what it is permitted to do; applied to arms control, that separation implies a treaty regime directed at capability faces an intractable inspection problem, since assessing how capable a system is would require access to algorithms, training data or model weights that no state will grant, while a regime directed at permission asks what a system is allowed to do, under what conditions, and with what record, which states can demonstrate without exposing the internals. The argument develops positions established in his own verified published work rather than introducing new ones. He has already published the verification questions this implies, asking which classes of action require fresh human authorization, under what evidence conditions permissions must narrow, what happens when communications, navigation or sensing degrade, whether a system may continue selecting or engaging after entering an uncertain environment, what must be true before higher permissions return, and whether an audit record can reconstruct what the system was allowed to do, what evidence supported that permission, and which human authority approved the boundary. He has also published the reason these are tractable, that they require a system to demonstrate that certain transitions are allowed and others impossible under declared conditions, which is testable without inspecting classified internals. The piece extends his related argument that narrow proofs through cryptographic attestation, tamper-evident logs and standardised audit records change the object of negotiation from a request to be trusted into a demonstration about the specific rule agreed, and his proposal that states declare authority thresholds in machine-readable form so that a shared system defaults to the most restrictive declared threshold. The significance for this record is that it converts his architecture into an arms-control proposition. Where his engineering work specifies how authority should be bounded, recorded and made to expire, this states why those same properties are the ones a verification regime can actually check, which is the bridge between his technical programme and the diplomatic process it addresses. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no treaties, negotiating bodies, states or officials are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 3, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “The AI Scientist Has a Metrology Problem,” his fifty-first contribution to the outlet and the twelfth in a sequence on machine-assisted scientific inference. The thesis stated in the title identifies the layer beneath the rest of that sequence. Its preceding pieces addressed what an automated system may claim, when it must abstain, how far a search extends, when to stop looking, and how a detection differs from an explanation; this one addresses the measurements those judgements rest on, and holds that the discipline of measurement itself is the missing requirement. The argument draws together positions verified in his own published work: that every measurement has a physical limit no amount of processing can move, which he set out in his analysis of information bounds in scientific systems; that a claim must travel with the conditions under which it was obtained, including the instrument state and the validated operating domain, which he has published for navigation integrity, for defence acquisition and for cryptographic assurance; and that a confidence figure describes a model’s output under its own assumptions rather than establishing that the underlying measurement supported the displayed detail, which is the formulation he used in his work on reconstructed imagery. Applied to automated discovery, the requirement is that uncertainty, calibration status and the conditions under which an instrument was validated be carried with a result rather than assumed by whatever consumes it. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no instruments, laboratories, standards bodies or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 3, 2026 · Authored by Burak Oktenli · Fifteenth placement with the outlet

The Washington Examiner published the applicant’s op-ed “Trump Promised to Drain the Swamp. AI Is Making It Unfireable,” his fifteenth placement with the outlet. The argument moves his authority research out of defence entirely and into domestic administrative governance, and it is the clearest instance to date of him arguing a position from inside the political premises of his readership rather than against them. He begins from the government’s own published figures rather than from an assertion: a 2025 budget-office directive told every agency to accelerate adoption and remove barriers, and the consolidated federal inventory as of April 2026 lists several thousand individually reported use cases across the executive branch, of which around half are deployed or piloted and several hundred are classified BY THE AGENCIES THEMSELVES as high-impact, meaning capable of shaping a decision with legal or material consequences for a citizen. His observation about them is the aggregation point he has developed elsewhere: taken one at a time almost every entry is modest, a model that flags a claim, a workflow that routes an application, a tool that drafts a determination for a human to sign, and none of them looks like a bureaucracy, while all of them together are becoming one. He names the gap directly, that an autonomous bureaucracy is being assembled from thousands of small pieces and its constitution has not been written. The analytical core tests that emerging system against four structural features of constitutional design and finds each one absent, not by intent but because each individual component is too small to appear to need it. SEPARATION OF POWERS: in many automated chains the same pipeline flags a case, scores it, decides it and triggers enforcement, with a human who sees a recommendation rather than a choice, so the rule-writer, the applier and the judge have collapsed into one system. APPEAL: a citizen can appeal to a human, but if that human’s only tool is to run the same workflow again the appeal is theatre, and a real appeal requires a person with the authority and the information to overrule the machine. JURISDICTION: automated decisions increasingly run across agencies, so an identity check in one system feeds an eligibility model in another which feeds a payment decision in a third, and when the outcome is wrong no single agency owns the chain while the citizen is passed between offices each correctly saying the error did not originate with it. ACCOUNTABILITY: when a caseworker denies a benefit there is a name on the letter, and when a model does it there is a vendor product, an agency dataset and a reporting category, so there is no one to hold responsible and therefore no one with an incentive to notice. He grounds the risk in two documented public-administration failures rather than in hypotheticals, a state-level automated fraud-detection system that ran without human review for nearly two years and was later found to have an error rate above ninety per cent, and a national programme in another country that a royal commission subsequently found to be neither fair nor legal and which ended in the refunding of hundreds of thousands of debts. His reading of both is structural rather than moral: malice had nothing to do with it, and the systems kept running because nobody had jurisdiction to stop them, nobody was accountable for their output, and the appeal path led back to the machine. *** The argument he then makes to his own readership is the notable one. *** He observes that a civil servant can be reassigned, disciplined or removed, while a workflow cannot be fired and can only be switched off by someone with the authority who knows where the switch is, so an administration that has worked to make the bureaucracy answerable should not replace it with one that answers to no one. He adds the durability point, that whatever is built now will be inherited and operated by the next administration, and that the workflows will not care who won the election, concluding that anyone who would not accept an unaccountable human agency should be the last to accept an unaccountable machine one, because the machine version outlasts every political appointee who approved it. This is his authority-expiry and named-ownership architecture argued in constitutional rather than engineering language, to an audience that already holds the premises it depends on. His four proposed rules belong in statute rather than in guidance a later administration can rewrite: every automated decision with legal or material effect on a person bears the name of the official who owns it, not the vendor and not the agency but a person; every such decision carries a right of appeal to a human who can overrule the system rather than only rerun it and who can see the evidence it used; every chain crossing agency lines has a single owning agency with jurisdiction over the whole chain so the citizen has one door to knock on; and automated authorities EXPIRE, requiring reauthorization on a fixed schedule as a spending programme does, so that the default is that a workflow stops unless someone with a name decides it should continue. That fourth rule is the permission-lapse structure from his formal work stated as administrative law. He closes on the observation that the founders wrote a constitution before trusting the government it created, and that the principle has not changed because the government now runs on software.

Read on the Washington Examiner ↗ Author page ↗

Published Commentary September 2026

One hundredth commentary placement, published by RealClearDefense

RealClearDefense · September 3, 2026 · Authored by Burak Oktenli · Ninth placement with the outlet, which maintains a standing author page for him · One hundredth original commentary placement

RealClearDefense published the applicant’s op-ed “The Pentagon’s Public Disclosures Need a Mosaic Test,” his ninth placement with that outlet and his one hundredth original commentary placement overall, subsequently distributed through the RealClearWire syndication network and republished by two regional broadcast stations in separate media markets, one of which has now carried two of his pieces. Its argument concerns the security consequences of cumulative public disclosure during a fast-moving campaign, and it follows the method that characterises his strongest policy writing: it locates the principle already present in existing doctrine, then proposes a procedural change rather than a new theory. He opens on the observation that information can change meaning when it is assembled, and grounds it in two existing frameworks rather than asserting it: departmental classification guidance explicitly provides for classification by compilation, illustrating how unclassified pieces can become classified in combination because the compilation reveals something more sensitive than any individual piece; and the operations-security programme makes the same point from the other direction, recognising that an adversary can piece together open-source indicators to estimate friendly intentions, capabilities or activities. His proposal is to make that logic a routine part of public campaign communications, which he names a MOSAIC TEST, and his governing image is that a safe tile can still change the picture. *** He then declines the conclusion the argument most easily invites. *** He states plainly that this is not an argument that too much is being disclosed, and lists what public information legitimately accomplishes: demonstrating progress, reassuring allies, supporting democratic accountability, countering false claims, and explaining why forces are being used. He also rejects the defensive response, arguing that classifying routine information merely because software can aggregate it would sacrifice speed, trust and useful transparency while creating more protected material without protecting what matters most, and that a military which says nothing creates its own strategic costs, since allies need confidence, citizens need evidence, and adversary propaganda should not be allowed to monopolise the first account of an operation. The narrowness of the ask is what makes it actionable. His illustration uses only figures the department itself released during a recent campaign, and his analytical point about them is that an adversary does not read a force count as a press-release writer does but compares it against deployment announcements, public imagery, contract data, replenishment schedules, maintenance reporting, budget documents, congressional testimony and previous statements, so that value emerges from the relationship among facts rather than from any single fact. He notes that the collection burden which once limited this kind of analysis is falling, as commercial imagery, searchable transcripts, procurement databases, shipping and aviation data and machine-assisted extraction turn scattered releases into a structured dataset. He reformulates the reviewer’s question accordingly: not whether a number is classified, but what estimate becomes easier once it is added to everything already public. The proposed test has four parts. A ROLLING DISCLOSURE LEDGER against which high-value operational releases are reviewed, tracking recently released force levels, readiness statements, operational statistics, logistics references, platform availability and weapons-use figures, whose purpose he is careful to define as preserving institutional memory of what the public record already permits an analyst to know rather than creating a new classified archive of public information. An ADVERSARY-VIEW RECONSTRUCTION in which a small red team is given only publicly available material and asked what a competent foreign analyst could estimate now that could not be estimated before this release, with publication proceeding if the answer is nothing material and the release delayed, generalised or separated from another disclosure if it materially sharpens an estimate of capacity, timing, limitation or dependency. TIME DECAY, since operational sensitivity expires and a detail risky during an active movement window may be harmless afterwards, so the test should favour delay and precision over permanent secrecy: protect the fact while it changes an adversary’s decision, then release it when that value expires. And TIERING, reserving the extra review for information touching operational tempo, force availability, munitions expenditure, logistics, deployment timing, maintenance or dependencies, on the stated ground that the goal is a better filter and not another clearance bureaucracy. The piece belongs to the same family as his work on corroboration and provenance: there he argued that agreeing feeds inheriting a common source do not supply independent evidence and that one inferred detail can propagate until repetition makes it look established; here the same aggregation logic is turned outward, examining what an adversary can assemble from separately harmless disclosures. His closing formulation states the procedural change in one line, that the department already asks whether a fact is classified, releasable and operationally sensitive, and the next question should be what that fact becomes after it joins the facts released yesterday.

Read on RealClearDefense ↗ RealClearDefense author page ↗ Broadcast-group republication ↗ Second broadcast republication ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 3, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “The Best Discovery Machines Know How to Disappoint Us” under its Analysis designation, the twenty-third of his contributions to carry that label, his fiftieth for the outlet, and the eleventh in a sequence on machine-assisted scientific inference developed since mid-August. The thesis stated in the title inverts the usual measure of an automated discovery system: the capacity to return nothing is evidence of quality rather than of failure, because a system that always finds something has not demonstrated that it can discriminate. It consolidates positions established across that sequence in his own verified published work, that an unresolved state must remain a legitimate output rather than being forced into a known class, that a null result closing a possibility is a scientific product and not a failure, that the extent of an automated search belongs in the claim that search produces, and that a standard for declaring a result is incomplete without a rule governing when to stop looking. It also restates, in a scientific register, the requirement he has published across his defence and governance writing, that a system report the evidentiary condition of its own output rather than a score that conceals it. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no cases, disciplines, methods or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 2, 2026 · Authored by Burak Oktenli · Forty-ninth contribution to the outlet

Eurasia Review published the applicant’s opinion piece “The AI Age Needs a New Definition of Sovereignty,” his forty-ninth contribution to that outlet since the relationship began. The thesis stated in the title argues that the concept requires redefinition for systems of this kind, and it develops a position established in his own verified published work rather than introducing a new one. That position holds that sovereignty over such systems is an operational capacity rather than a matter of nominal ownership: a state unable to retrain a model without foreign cloud infrastructure, patch it without a proprietary toolchain, validate a new version without external support, or move the workload when access conditions change holds less of it than a capability comparison suggests, and vendor dependence therefore becomes a form of strategic dependence. He has published the supporting enumeration in the same terms, listing compute, energy, data pipelines, integration, cyber assurance, retraining, model monitoring, operator support, software updates, export restrictions and the cost of leaving a supplier as the factors that sit offstage during a demonstration while determining whether a capability remains usable afterwards. The piece also extends his argument that a second geography is forming alongside the first, a geography of trusted access created by research-security rules, data protection and export controls, in which security alignment increasingly affects the transaction cost of building and fielding technology. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no states, agreements, companies or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · September 2, 2026 · Authored by Burak Oktenli · Fourteenth placement with the outlet, which maintains a standing author page for him

The Washington Examiner published the applicant’s op-ed “Trump’s AI Mandate: Defining Killer Robots Won’t Save Us — Hardcoding Human Control Will,” his fourteenth placement with the outlet, subsequently syndicated to MSN. It is timed to a policy window rather than to a news cycle: a presidential memorandum gave the defence department ninety days to update the directive governing autonomy in weapon systems, and he writes in the week that window closes, noting that the instruction was not simply to adopt artificial intelligence faster but to do so while preserving the chain of command and operational authority. His method here is the one that recurs in his strongest policy writing. He takes the leading proposal in the live debate, a major research institute’s recommendation to widen the definition of an autonomous weapon so that it reaches beyond drones and munitions to the software that fuses sensor feeds, selects targets and orchestrates the kill chain, and he ACCEPTS IT, stating that if software performs the function that makes a weapon autonomous then policy should not stop at the airframe. Having granted the point he identifies what it cannot reach: a definition tells the military what falls under a rule, but it does not tell commanders what authority humans must retain once a system falls inside it. He then credits the existing directive’s requirements as strong foundations, listing its demands for appropriate levels of human judgment over the use of force, rigorous testing, transparent system status, and systems that terminate an engagement or seek additional operator input when they cannot continue within the commander’s intended constraints. His three proposed upgrades are his research programme translated into directive language. The first is to specify HUMAN AUTHORITY rather than the amount of human involvement, identifying for a given system and mission which decisions remain reserved to a person, under what conditions the system may act without additional input, and which boundaries it may never cross on its own, which he names an AUTHORITY ENVELOPE, a mission-specific statement of what the machine may do and what remains a human decision. The second is that the boundary be enforced BEFORE ACTION, so that where a decision is reserved to a human the system is technically unable to cross that line until the required authorization has occurred, supported by his sharpest formulation in the piece: a safeguard discovered only in an after-action review was not a safeguard at the moment it mattered. The third is that human control be VERIFIABLE RATHER THAN ASSERTED, with the system preserving enough of the decision record to establish what information was presented, what the machine recommended, what the human authorized, what operational limits were active, and in what sequence those events occurred, with the explicit qualification that the goal is not to automate legal judgment but to preserve evidence that the human role required by policy actually existed in practice. He is careful that none of this asks the department to slow adoption, grounding the proposals in principles the department has already embraced through its own artificial-intelligence ethics framework, which calls for systems to be traceable, reliable and governable including the ability to disengage or deactivate systems behaving in unintended ways, and framing the update as translating those principles into testable requirements. His closing argument is about tempo and is the through-line of his published work: a person can remain formally present while the decision window becomes too short to understand the evidence, challenge the recommendation, or exercise a reserved authority before the action is complete, so the answer is not to demand a human click at every millisecond but to decide in advance which decisions must remain human and to engineer the system so those decisions cannot silently migrate to software as tempo increases. He states the aim of the update as telling commanders what decisions remain theirs, engineers what boundaries must be enforced, and testers what evidence must exist to show that human authority survived machine speed, and ends on the line the piece is built toward, that the point is not simply to define the autonomous weapon but to make human command technically real.

Read on the Washington Examiner ↗ Author page ↗ MSN syndication ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 2, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “Industrial AI Needs an Authority Test Before It Touches the Physical World” under its Analysis designation, the twenty-second of his contributions to carry that label and his forty-eighth for the outlet. The thesis stated in the title moves his central requirement into industrial and manufacturing control, a setting distinct from the defence, scientific and alliance registers in which he has developed it. Its claim is that the decisive question for a model placed on a production line is not how accurate it is but what it is permitted to do once its accuracy no longer holds, which is his standing separation of capability from authority applied where the consequence is physical rather than informational. The distinction he draws between a recommendation and an actuation is the operative one: a wrong recommendation costs a review cycle, while a wrong command moves a physical thing, so the two require different permissions even when produced by the same model. The proposed authority test is his published structure carried into a new domain, asking before deployment rather than after an incident which actions the system may take without human confirmation, under what sensor and calibration conditions those permissions remain valid, what the system does when those conditions fail, and whether the resulting record allows an investigator to reconstruct which permission was in force and who set it. Those four questions restate, for industrial control, the requirements he has published for federal artificial-intelligence deployment, for allied autonomous systems, and for cryptographic assurance: authority bounded per action class, permissions that contract as their supporting evidence deteriorates, degradation behaviour specified in advance, and a reconstructable record of the authority under which a system acted. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no plants, vendors, standards, incidents or figures are attributed.

Read on Eurasia Review ↗

Published Commentary September 2026

Op-ed published in Eurasia Review

Eurasia Review · September 1, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Calling AI the New Columbus Is Too Accurate to Be a Compliment,” his forty-seventh contribution to the outlet. The thesis stated in the title takes a comparison usually offered as praise and argues that it holds more closely than those using it intend, so that its accuracy is the problem rather than the compliment. The move is one he has made before in his published work: examining what a borrowed vocabulary carries with it rather than accepting the framing it arrives in. It extends two positions verified in his own text. The first is his caution that scientific and historical vocabulary can become more powerful in public discussion than the thing it originally described, which he stated in his analysis of causal-order claims and applies here to a metaphor rather than a technical term. The second is his argument that discovery is a status conferred when evidence meets a stated standard, not a claim established by arriving somewhere first and naming it, which he set out in his sequence on machine-assisted scientific inference. This entry is held to the thesis stated in the title and to those verified prior positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no historical, political or attributional claims from the piece are recorded here.

Read on Eurasia Review ↗

Published Commentary September 2026

Memorandum published by a British foreign-policy institute

Britain’s World, the publication of the Council on Geostrategy · The Memorandum No. 43.2026 · September 1, 2026 · Twenty-first outlet to publish his work

Britain’s World, the publication of the Council on Geostrategy, a London foreign-policy institute, published the applicant’s piece “Britain Settled the Algorithm Question, But the Assurance Question Is Open” as a numbered entry in its Memorandum series, his first placement there and the twenty-first publication to carry his work. The format is a policy memorandum addressed to a national audience rather than a general-interest commentary, and the piece is written to that register throughout, accepting an existing national position and identifying the work it leaves undone rather than contesting it. He takes as settled the national cyber authority’s published migration timetable, its decision not to support quantum key distribution for government or military use, its designation of post-quantum cryptography as the best available mitigation, and its judgement that a quantum-distribution implementation alone is insufficient evidence that data in transit is secure. His argument is that this last judgement should become an ACQUISITION DISCIPLINE rather than remain a technical opinion. Choosing an algorithm answers a cryptographic question; assuring a deployed system requires evidence about authentication, key lifecycle, control software, timing, relays, fallback behaviour, operator interfaces, and the way those elements fail together. He notes that the authority’s own guidance supports the point by describing algorithm standards as building blocks from which protocols, products and services must still be constructed, and identifying areas such as industrial-control protocols and web public-key infrastructure where migration cannot be reduced to swapping a primitive, and he locates the value of an assurance case precisely there: it makes surrounding dependencies and failure behaviour visible BEFORE AN APPROVED ALGORITHM IS MISTAKEN FOR AN APPROVED SYSTEM. He argues the country retains an interest in evaluating a technology it has chosen not to deploy, since its networks and allied systems will encounter products described as quantum-secured regardless, and observes that the label itself is losing usefulness as architectures diversify, because a vendor may apply it to distribution, post-quantum, or hybrid designs with different trust assumptions and failure modes, so procurement must translate it into a specific claim: what is protected, against what threat, under what operating conditions, and with what dependencies inside the assurance boundary. He credits the physics as a strong but bounded claim, then sets out TWO LINKED ATTACK SURFACES, the classical and operational one comprising authentication services and credentials, key-management software, relay facilities, network orchestration, operator consoles and the logic deciding what happens when a link degrades, and the implementation one comprising detector behaviour, source imperfections, calibration, timing side channels and optical injection, noting that published national-agency work has demonstrated attacks on commercial equipment. His formulation of the failure is the one that carries: a security claim fails if either surface falls outside the evidence, so a commander could be told, accurately, that the link is secure while the conversation as a whole remains vulnerable elsewhere in the system. He applies the same test to the route the country did choose, observing that a cryptographic inventory can identify which algorithms are present but cannot establish whether key management is compromised, whether a transitional connection falls back into an unexpected mode, whether an operator is warned when protection changes, or whether an implementation behaves safely under denial, and concludes that ALGORITHM APPROVAL AND SYSTEM ASSURANCE ARE SEPARATE EVIDENCE PROBLEMS even when the algorithm choice is sound. His proposed artefact is a quantum communications assurance case in four parts: state the security claim and the trust boundary, naming which property the cryptography provides, where authentication originates, who manages keys across their lifecycle, which relay nodes are trusted, and whether timing, calibration, control software and operator displays sit inside the assessed boundary; name and test classical and quantum dependencies together, carrying evidence from the cryptographic primitive through to the system built around it; specify denial and downgrade behaviour before deployment, with a pre-approved response of halting, failing over to a defined mode, or continuing under explicitly accepted risk, and an operator display showing which state is active and WHO ACCEPTED THE RESIDUAL RISK; and adversarially evaluate the whole system, attacking authentication, key management, control software, trusted nodes, timing and calibration paths, detectors and optical interfaces, fallback logic, and the consoles that report system state, rather than measuring link performance alone. He frames these as an evidence gate that preserves research and migration rather than constraining them, with security credit following only when both dependency classes sit inside one evidence package and have been tested as a system, and he closes on the memorandum’s governing line: the difficult choice about cryptography has been made, and the next task is to make the security claim auditable. The four requirements restate, in a cryptographic setting, the position he has developed across his published work: that a claim must carry the conditions under which it holds, that behaviour at the point of degradation must be specified in advance rather than discovered, that the operator interface must show the current state and its owner, and that evaluation must address the system rather than the component.

Read the memorandum ↗

Published Commentary September 2026

Article published by an allied army’s official professional development platform

The Cove · September 1, 2026 · Authored by Burak Oktenli · Twentieth outlet to publish his work

The Cove, the professional development platform of the Australian Army, published the applicant’s article “The Loudest Alarm Is Usually the Wrong One,” his first placement there and the twentieth publication to carry his work. The venue is materially different from a magazine or a news outlet: it is a service-run platform for the professional education of a national army, hosted on that country’s defence domain, and its articles are structured for practitioner use, with a summary, an estimated reading time, corps and rank tagging, reader reaction controls, and discussion questions appended for use in professional military education. Three such questions accompany the applicant’s piece, asking readers how they would decide which information is genuinely important to a mission, which decisions they are comfortable allowing a system to make about what they never see, and what follows if personnel become accustomed to treating the highest-priority alert as automatically the most important problem. The platform carries the standard disclaimer that views are the author’s own and do not represent the position of that army, its defence department or its government, and this record makes no claim of endorsement or affiliation. His argument is that ranking automated alerts by anomaly score is, for operational purposes, close to backwards, because a statistical departure and an operational consequence are different quantities: a score measures how unusual a reading is, while mission impact measures what would follow if the condition behind it were real. A large, attention-grabbing deviation in a subsystem the mission does not depend on can matter far less than a small, quiet drift in one the mission rests on, so single-axis ranking sinks the dangerous case to the bottom of the queue and puts the harmless one at the top under a red banner, spending the scarcest resource in the monitoring chain, the operator’s attention, exactly where it is needed least. He grounds this in an official accident investigation into a warship collision in which control had been transferred between stations without the helmsman realising, noting the investigators’ finding that the information which mattered most was present in the system throughout and was never lost, only never absorbed, and that the design of the control interface itself increased the likelihood of the operator errors involved. He is careful that this is not an argument against anomaly detection but for treating it as one input rather than the whole answer, and proposes ranking on two axes, statistical strength and mission consequence, observing that the resulting quadrants are asymmetric and that the case any worthwhile scheme exists to protect is the LOW-SCORE ANOMALY IN A HIGH-IMPACT SYSTEM, precisely because single-axis ranking would have discarded it. His harder point is that mission impact is not a fixed property that can be computed once and stored, since what a subsystem contributes changes with phase, configuration and tasking, so a sensor irrelevant in transit becomes critical in a terminal engagement and a redundant communications path becomes a single point of failure in another posture; a scheme with impact weights frozen at design time will rank correctly in the phase it was tuned for and misrank in every other, becoming quietly wrong at exactly the moments that matter. From this he draws requirements that restate his standing position in a new setting: the mapping from subsystem to mission consequence is a modelling choice that should be made by the operators and engineers who know the platform rather than inferred silently by a model optimising whatever was in its training data; the basis for each impact estimate should travel with the alert so an operator can see why something ranked where it did and override it when the model’s assumptions do not fit; and a ranking that cannot be inspected cannot be trusted, because nobody can catch it when a confident but wrong estimate sends scarce attention to the wrong place. He notes that none of this requires new technology, only a design choice, and closes on the observation that the systems which win the competition for human attention should be the ones that matter rather than the ones that shout.

Read on The Cove ↗

Published Commentary September 2026

Op-ed published in Eurasia Review, replying to a writer who had cited him

Eurasia Review · September 1, 2026 · Authored by Burak Oktenli · First piece published there that day

Eurasia Review published the applicant’s opinion piece “The Next Autonomous-Weapon Failure May Be a Drone That Obeys Every Rule,” his forty-fifth contribution to the outlet and the first he published there that day. It is notable in this record for CLOSING A CITATION LOOP: the syndicated feature recorded elsewhere on this page, which quoted his Articles of War analysis on where responsibility for an autonomous system’s actions properly sits, is answered here directly. He names its author, credits her argument that lethal systems are outrunning the institutions expected to hold humans accountable, restates what she took from him accurately, and then advances the position rather than merely accepting the compliment, writing that she is right that accountability begins earlier than the instant a weapon strikes and that the next step is to make that insight executable. A third party used his work; he replied in print and extended it. His argument sets aside the rogue-machine framing as the least useful image in the debate. A weapon does not need to rebel to escape meaningful human control; it may only need to follow its instructions perfectly after the conditions that justified them have disappeared. He illustrates with a mission in which the communications link drops, navigation confidence deteriorates, and obscuration or deception degrades the sensor picture, so the software switches to onboard guidance and continues exactly as designed: nothing has escaped, nothing has acquired a motive, the system may be functioning precisely as specified, and the governance question is whether it should still possess the same authority. He then distinguishes degrees of delegation that are usually collapsed together, from recommending a target, to ranking targets, to selecting an object, to completing an engagement after losing contact with the responsible human authority, adding that a swarm creates a further layer because a lethal outcome may emerge from interaction among several machines rather than one identifiable decision point. The design principle he states is that PERMISSION SHOULD FOLLOW EVIDENCE, so authority contracts step by step as the link is lost, as navigation integrity falls below its validated threshold, as sensors disagree, and as conditions move outside the validated operating domain, which he contrasts with a single binary switch marked autonomous or not, on the ground that authority should be dynamic because the evidence supporting it is dynamic. He credits an existing national directive with recognising part of this logic in requiring appropriate levels of human judgment, realistic-condition performance, and operation within stated time, geographic, environmental and operational parameters, while arguing the engineering principle should be made unmistakable, and states the failure mode in one line: the loss of evidence should never silently produce the inheritance of authority. His second principle is an asymmetry, CONTRACT QUICKLY AND RECOVER DELIBERATELY. If a link returns after thirty seconds, full authority should not return with it, because one restored signal does not establish that the conditions supporting the original permission have been restored; recovery should require pre-defined evidence such as multiple independent sensor confirmations, re-established navigation integrity, authenticated communications or renewed human authorisation, scaled to the consequence. Permission should be easy to lose when its supporting evidence fails and harder to regain. He also dismantles the phrase human in the loop by asking whether the operator has enough time, can examine the evidence beneath a recommendation, can see uncertainty rather than a confidence score, can determine whether apparently corroborating sensors depend on the same underlying source, and can stop the process without an operational penalty large enough to make approval the expected default. Where those answers are no, the human is present while substantive control has migrated elsewhere. The contribution to the diplomatic debate is his conversion of a normative principle into verification questions that do not require inspecting classified algorithms or model weights: which classes of action require fresh human authorisation, under what evidence conditions permissions must narrow, what happens when communications, navigation or sensing degrade, whether a system may continue selecting or engaging after entering an uncertain environment, what must be true before higher permissions return, and whether an audit record can reconstruct what the system was allowed to do, what evidence supported that permission, and which human authority approved the boundary. He argues these are testable, requiring a system to demonstrate that certain transitions are allowed and others impossible under declared conditions, which makes authority architecture relevant to arms-control verification and not only to design. He rejects both familiar slogans as insufficiently precise, and closes on the observation that a future failure may involve no malfunction at all: the machine follows every instruction, its sensors degrade as anticipated, its fallback activates correctly, it remains entirely inside the constraints humans gave it, and civilians are still harmed because those constraints allowed too much authority to survive the disappearance of the evidence supporting them.

Read on Eurasia Review ↗

Published Commentary September 2026

Analysis published in Eurasia Review

Eurasia Review · September 1, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “AI Is Giving the Kill Chain a False-Precision Problem” under its Analysis designation, the twenty-first of his contributions to carry that label and his forty-sixth for the outlet. He opens on a targeting cell receiving a degraded frame in which a reconstruction layer denoises and sharpens an image that the raw sensor did not clearly resolve, and states the governing question: not whether the enhanced image looks plausible, but whether the sensor ever measured the detail that now appears decisive. He establishes the operational stakes from programmes of record rather than speculation, citing a defence ministry’s battlefield platform analysing over one hundred thousand video streams monthly against a dataset of five million annotated frames, an alliance targeting project seeking to cut information-to-decision latency by more than half, a national investment exceeding one billion pounds in a digital targeting web, and an army programme ingesting space, high-altitude, aerial and terrestrial sensor data. He then does something the record should note: he EXPLICITLY DISCLAIMS the stronger inference those examples might invite, writing that none of them proves generative reconstruction is being used for lethal targeting judgments and that it would be irresponsible to imply otherwise, and resting his case instead on the more basic fact that algorithmically processed sensor data is entering decision chains at greater scale, from more sources, and at higher speed. The technical core is that reconstruction is an inverse problem in which a degraded measurement does not correspond to one unique scene, so several materially different scenes can fit the same observation and the method must fall back on learned priors, regularisation, training data or shape assumptions. He is careful that this is not automatically a flaw, since every reconstruction method uses assumptions and modern methods recover useful structure. The governance failure begins where model-supplied specificity becomes visually indistinguishable from sensor-derived evidence, and he supports it with published findings: work on generative super-resolution identifying realistic-looking details unsupported by the input, computational-imaging research showing non-physical predictions beyond the conditions where the inverse mapping is recoverable, and documented cases in medical imaging where reconstruction created plausible but false anatomical features. He names the failure FALSE PRECISION, a narrow, confident-looking product produced from evidence that still supports meaningful alternatives, and identifies the dangerous case precisely: not an absurd hallucination, which invites scepticism, but a plausible discriminating feature the model supplies because it fits the learned prior. His second move is the one that connects to his research programme: inside a research notebook false precision is an uncertainty problem, while inside a military system it becomes a PROVENANCE problem, because at each handoff the receiving people and machines move farther from the original sensor until a reconstructed detail crosses an organisational or national boundary and arrives looking like observed fact. He extends an alliance data-quality framework’s recognition of lineage into what he calls EPISTEMIC PROVENANCE, not only where a product came from but which parts of its apparent specificity were measured and which inferred, and he links it explicitly to his own earlier published argument that agreeing feeds inheriting a common upstream source constitute a corroboration trap, naming false precision its complement: one inferred detail propagating through several products until repetition makes it look independently established, so that the chain has not discovered more evidence but only moved the same inference farther from its origin. His four requirements are a PROVENANCE OF SHARPNESS preserved in machine-readable form and surviving every copy, fusion, model handoff and coalition transfer, with the explicit note that a single confidence score is insufficient because it describes the model’s output under its own assumptions rather than proving the measurement contained the displayed detail; exposure of material alternatives where several reconstructions are consistent with the data and would lead to different judgments, rather than forcing one photogenic winner; AMBIGUITY TESTING that asks not whether a model usually gets the answer right but whether it recognises when the evidence does not justify a unique answer, with a specified test battery and a pass condition covering whether confidence falls appropriately and whether downstream software preserves the uncertainty; and unresolved output able to generate a new collection requirement, so that the architecture makes it easier to ask for evidence than to manufacture specificity. He closes on the coalition dimension, proposing that evidentiary lineage become part of interoperability through a minimal shared schema preserving the sensor-to-inference boundary, since allied architectures need not share algorithms but do need a common way to say which feature was measured, which was inferred, and which uncertainty remains unresolved, and he states the standard in one line: commanders do not need the sharpest image a model can produce, but the sharpest image the evidence can defend.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 31, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “We Can See Earth’s Past, We Cannot Send It a Message,” his forty-fourth contribution to the outlet. Its method is the one that recurs across his scientific writing: take a proposal that sounds plausible for a few seconds, identify precisely which step fails, and then ask what survives the failure. He separates two questions routinely conflated, recovering information that was already emitted in the past, which is ordinary physics, and sending newly created information to an event that has already occurred, for which established physics supplies no mechanism. The first he grounds in the delay inherent in every astronomical observation and in the light echo, where part of a stellar outburst travels directly to an observer while another part scatters from surrounding dust and arrives later, and he isolates the point most easily missed: the scattered light has not travelled backward through time but forward for longer. He then turns the problem around, noting that electromagnetic radiation which left the planet decades ago did not cease to exist and continues outward, so that if some of it later interacted with suitably positioned matter and a fraction were redirected back, it could be received long after emission. That, he argues, would not be retrocausal communication but ASTRONOMICAL ARCHAEOLOGY, the universe returning information the planet itself placed into its own future light cone. He observes that a modern astrometric catalogue containing positions and motions for nearly two billion sources does not supply a convenient reflector but does something more useful, converting a geometrical question that once belonged to speculation into one that can be confronted with real catalogues, which he describes as how a strange idea becomes a research question. The refutation is causal rather than technical: a message first exists at the moment it is chosen, a transmitter can send it to a later event, a reflector can pass it to another later event, and a computer of any kind can process information only after that information is available to it, so adding stages changes direction, wavelength, encoding, power or trajectory without making the destination occur before the selection. A mirror reverses a beam’s spatial direction, not the ordering of cause and effect, and an optimiser can find the best route available without creating a route the causal structure does not provide. He closes the same door on quantum arguments, noting that the standard teleportation protocol still requires classical information to reach the receiver before the state can be reconstructed, and setting a strict test for any effect described with temporally suggestive vocabulary: can an experimenter choose a classical message now and cause that choice to become accessible at an earlier event? If not, whatever occurred is not communication with the past. He adds a caution that belongs to his wider argument about claims, that scientific vocabulary can become more powerful in public discussion than the physics it originally described. Having established the prohibition he treats it as productive rather than terminal, listing the lawful problems that remain, including long-range optical communication, machine-assisted pointing and relay scheduling, photon-counting sensitivity, reconstruction of faint delayed radiation by distributed telescopes, and archives deliberately created to be recovered decades later, and he anchors the direction in a completed deep-space optical demonstration that transmitted high-definition video across tens of millions of kilometres and later operated across hundreds of millions. The formulation that carries beyond the subject is his distinction between limitations that are optimisation problems and limitations that define the set over which optimisation is permitted to operate, with the corollary that causal order is not another tunable engineering parameter. That is the same distinction he drew between a physical ceiling and an engineering inefficiency in his work on exotic propulsion and on tuning against physical limits, applied here to the boundary between the physically forbidden and the merely difficult.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published by The Defence Horizon Journal, a new outlet

The Defence Horizon Journal · August 31, 2026 · Authored by Burak Oktenli · Nineteenth outlet to publish his work · Standing author page

The Defence Horizon Journal, a defence publication that operates an editorial office and a peer board and maintains a standing author page for him, published the applicant’s analysis “The Alliance Has No Common Answer to a Machine-Speed Question,” his first placement with that outlet and the nineteenth publication to carry his work. The piece is structured to the journal’s format, opening with an abstract, an explicit problem statement, and a “so what” section stating the recommended action. Its argument is the applicant’s central research claim applied to alliance governance. He credits the alliance with three years of work on the technical scaffolding for fielding autonomous systems across member states, including an autonomy implementation plan, principles of responsible use, and a cadence of interoperability exercises, and grants that it has largely succeeded at interoperability understood as an engineering property, meaning allied machines can operate as one. His claim is that it has not resolved a harder property, authority: when an autonomous system proposes a lethal action faster than a human can meaningfully review it, and two allies operating the same kill chain hold different national thresholds for permitting that action, whose threshold governs, and who is accountable when they disagree. He grounds the divergence in documented policy rather than hypothesis, citing one member state’s coalition agreement rejecting systems that operate entirely without human control alongside another’s directive permitting them subject to appropriate levels of human judgment, a standard its own guidance describes as deliberately flexible, and noting that the allies do not share a vocabulary, since expert bodies have worked in terms of meaningful human control while one member rejects the phrase. He illustrates the mechanism with a distributed strike sequence in which a loitering munition selects and strikes independently, a human-confirmed follow-up occurs, and a decoy draws air defences that another autonomous system destroys, observing that the chain works precisely because its timing is compressed below a deliberate human decision cycle, so a shared network does not pause at a national boundary to ask which standard applies. The distinction he draws is between adversary-facing escalation through misperception and what he names ESCALATION BY DIFFUSION within an alliance, in which an action taken at one member’s permissive threshold on a shared system commits the coalition to consequences its more cautious members never authorised, so that the cautious ally discovers after the fact that it was party to a strike its own rules would have forbidden, because the system was interoperable and the authority was not. He anticipates the obvious objection, that the chain of command resolves this, and answers that machine-speed autonomy is precisely the condition under which the commander’s deliberate decision has been compressed out, noting that allied joint targeting doctrine is a phased process built around commander decision points, which is the tempo autonomous engagement is designed to outrun. He also identifies three plausible loci of command responsibility that are not designated in advance: the state whose platform executed the strike, the state whose sensors cued it, and the commander whose network configuration fixed the engagement threshold. His three commitments require no new institutions: treat authority as a graded, declared property, so that before fielding an interoperable capability each state declares in machine-readable form the threshold at which its rules permit autonomous lethal action and the shared system defaults automatically to the most restrictive declared threshold, on the principle that an alliance is only as permissive as its most cautious member consents to be; make every engagement leave an authority record logging at the moment of action which threshold applied and under whose authority, noting that the technology exists while the requirement to use it does not; and treat threshold disagreement as a planning input rather than an operational surprise, so that coalition exercises rehearse what happens when allied authorities do not align as well as whether allied systems interoperate. He proposes that the standardisation machinery already certifying technical interoperability be extended to certify authority interoperability, making a declared threshold, an authority record, and a rehearsed disagreement procedure as much a condition of fielding on allied networks as a common data format. The piece unites the two lines the applicant has developed all year: graded, declared, machine-enforceable authority with a tamper-evident record, and the alliance-assurance argument that technical interoperability can advance faster than reciprocal confidence in the authority rules the software enforces. His closing formulation, that an alliance able to make its autonomous systems act as one without agreeing whose finger is on the trigger or how fast it may move has built a machine-speed question it has not given itself a way to answer, states that union in a sentence.

Read on The Defence Horizon Journal ↗ Author page ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 28, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “COVID Had Quarantines and Vaccines. A Cyber Pandemic Could Infect the Cure.,” his forty-third contribution to the outlet. The thesis stated in the title identifies a disanalogy that the familiar public-health comparison tends to obscure: the countermeasures deployed against a biological outbreak sit outside the reach of the pathogen, whereas the machinery for detecting, coordinating and remediating a large-scale digital incident can share infrastructure, software lineage and dependencies with the thing it is responding to. The argument draws on three requirements established in his own verified published work. The first is his dependency argument, that channels sharing a common ancestor do not supply the independent evidence a naive count implies, since components inheriting the same reference or software family fail together, so agreement or availability cannot be assumed from apparent redundancy. The second is the continuity requirement he set out for allied technology access, that planning name pre-identified alternative systems, local fallbacks, degraded operating modes and data portability rather than assuming a service remains reachable. The third is his revocation requirement, that a legitimate decision to restrict, suspend or revert authority must be capable of becoming an actual change in system behaviour, on the principle that governance without an executable path to revocation is advice. The domain is cyber crisis response rather than autonomy, and the entry records it on that basis.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published by RealClearDefense

RealClearDefense · August 27, 2026 · Authored by Burak Oktenli

RealClearDefense published the applicant’s op-ed “A Quantum Advantage Is Not a Defense Capability Until It Survives the Whole System,” his eighth contribution to the outlet, which maintains a standing author page for him. He opens by naming a translation problem: physics papers can demonstrate a real advantage under carefully defined conditions, while defence organisations do not buy error exponents, witness values, or laboratory plots but systems. He takes quantum illumination as his worked example precisely because the science is both real and frequently overstated, citing the original theory showing that correlated signal-idler architectures could improve target discrimination in noisy and lossy regimes, later work quantifying a substantial error-exponent advantage under specific Gaussian-state assumptions, and experiments demonstrating correlation-enabled gains in optical and microwave systems, each referenced to its published source. His conclusion from that literature is stated carefully: those results justify serious research, and they do not establish that a deployable long-range quantum radar can defeat stealth aircraft at tactically useful ranges. He then examines the microwave demonstration most often invoked in the debate, noting that it used a short laboratory-scale configuration and a digital receiver and showed that the concept can improve binary target discrimination under particular conditions, while leaving unsolved military aperture, atmospheric loss, clutter, mobility, latency, countermeasures, target cross section, and long-range idler storage, and he frames this not as a reason to dismiss the technology but as a reason to measure the right thing. His prescription is a NO-FREE-ADVANTAGE RULE under which every claimed quantum gain is reported beside the total system cost required to produce it, illustrated in both directions: a theoretically meaningful improvement is not yet mission value if obtaining it requires a receiver that destroys bandwidth, cannot operate at useful temperature, imposes prohibitive storage loss, or demands size and power incompatible with the platform, while a modest gain may be operationally valuable where transmit energy must remain extremely low or where a different error mechanism gives commanders useful diversity, so the demand is for end-to-end gains rather than spectacular ones. From this he derives FOUR ACQUISITION GATES: physical advantage, whether the quantum resource beats an optimised classical physical baseline under matched constraints; integrated advantage, whether that gain survives real source, channel, detector, calibration, timing and processing losses; assurance advantage, whether the system can detect that the conditions supporting its quantum claim no longer hold and degrade safely; and mission advantage, whether the assured system improves a mission-relevant outcome enough to justify lifecycle cost and complexity, with the observation that a programme passing the first gate has demonstrated physics rather than a fieldable capability. He insists the classical baseline be the best relevant architecture under matched bandwidth, energy, aperture, observation time and false-alarm constraints, and for learning systems a tuned classical model with equivalent data partitions, leakage controls, compute budgets and operational metrics, warning that otherwise a programme can manufacture an advantage simply by choosing the wrong opponent. The third gate is where this piece connects most directly to his research programme, since he extends the same discipline to machine learning used inside quantum systems, crediting calibration, state reconstruction, syndrome decoding, anomaly detection and sensor fusion as promising because they are measurable, and then stating the constraint that a learned confidence score should not be allowed to hide a failing clock, invalid calibration, lost correlation, or a degraded sensor state, which is his standing requirement that a system report the evidentiary condition of its own output rather than a score that conceals it. He closes by arguing the approach protects quantum research rather than slowing it, since the most damaging outcome would be to convert legitimate laboratory progress into inflated operational promises, spend heavily against them, and then conclude the field failed when engineering reality arrives.

Read on RealClearDefense ↗ RealClearDefense author page ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 27, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “AI Can Discover New Materials Faster Than Science Can Validate Them” under its Analysis designation, the twentieth of his contributions to carry that label, his forty-second for the outlet, the second piece he published on that date, and the tenth in a sequence on machine-assisted scientific inference. The thesis stated in the title concerns the asymmetry between the rate at which candidates can be generated and the rate at which they can be confirmed. It applies the promotion boundary established across that sequence in his own verified published work: a system may propose candidates, rank hypotheses, quantify uncertainty and recommend what to examine next, while moving from a proposed candidate to an established result requires controlled measurement, discriminating tests, independent replication, and a theory that predicts something new, none of which faster generation can supply. It carries the related requirements he has published in the same sequence, that an unresolved status remain a legitimate output rather than being forced into a conclusion, that the extent of an automated search belongs in the claim that search produces, and that a declaration standard is incomplete without a rule governing when to stop. The domain is materials research rather than governance or assurance, and the entry records it on that basis.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 27, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “What the Next Collider Can Really Give the World Beyond Particle Physics,” his forty-first contribution to the outlet. The thesis stated in the title asks what a very large scientific instrument actually delivers beyond the science it is built to do. It applies the accounting discipline he has developed across his published work in several fields: that a claim of advantage or value is meaningful only against a stated, optimised and resource-matched comparison rather than against whatever alternative is nearest to hand; that an unmeasured term must remain visible as an unknown instead of inheriting full performance by silence, which was the error he named in examining exotic-propulsion claims; and that the conditions under which a result was obtained belong in the claim it supports. The subject is scientific-infrastructure investment rather than autonomy or assurance, and the entry records it on that basis.

Read on Eurasia Review ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 26, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “Ratcliffe’s Moscow Visit Highlights the Strategic Value of Intelligence Backchannels Across Eurasia” under its Analysis designation, the nineteenth of his contributions to carry that label, his fortieth for the outlet, and the fifth piece he published on that date, matching his highest single-day count at the outlet. The subject is intelligence diplomacy rather than technical governance, and the piece is notable in this record chiefly for the evidentiary discipline it applies to a story built on an undisclosed agenda. He states repeatedly and explicitly what the public record does and does not establish, declining to present the visit as evidence of any particular mission, noting that a senior official’s public characterisation narrows the plausible framing without revealing the substance, and observing that the temptation is to turn an undisclosed agenda into a definitive story. From the verifiable facts he draws a narrower conclusion: whatever the agenda, the channel exists and is treated as potentially valuable. His analytical frame is that intelligence relationships matter not because officials can secretly negotiate settlements but because, where formal diplomacy is weak, such channels can carry warnings, clarify intentions, test whether an apparent opening is genuine, and reduce the chance that ambiguous actions become uncontrolled escalation, a function he notes is identified in published legislative research on foreign intelligence relationships. He distinguishes deconfliction from negotiation, treating practical measures around a delegation’s physical presence as risk management rather than evidence of political bargaining, and sets a role-clarity principle: use such channels to improve strategic understanding while keeping consequential policy decisions inside established diplomatic and allied processes, since the strongest backchannels reinforce the wider decision architecture rather than compete with it. He grounds the argument in precedent rather than assertion, examining an earlier mission by a predecessor that did not prevent the event it warned against, and drawing the limit precisely: a channel can transmit information accurately but cannot force the recipient to change policy, though it can still remove one form of uncertainty by making a consequential message difficult to misread. He extends the analysis to other theatres and to the strategic-stability environment following the expiry of the last bilateral treaty limiting deployed strategic forces, arguing that stability is becoming multipolar and that the problem is not simply more weapons but fewer mature risk-reduction mechanisms among some states that possess them, so high-risk relationships need redundant channels as decision timelines shorten. Two formulations connect this piece to his technical work. The first is his statement that the objective is not trust but controlled uncertainty, which is the same position he argues in verification, where the aim is to reduce the amount of trust a claim requires rather than to manufacture trust. The second is his observation that the most useful outcome of such a contact may be a reliable negative answer, which is the argument he makes for scientific practice, that a null result closing a possibility is a product rather than a failure. Applying both to intelligence diplomacy, in a piece that refuses to over-read its own subject, demonstrates the same epistemic standard operating outside his technical field.

Read on Eurasia Review ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 26, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “Bill Gates Is Right About AI Governance: Institutions Need Systems They Can Actually Govern” under its Analysis designation, the eighteenth of his contributions to carry that label, his thirty-ninth for the outlet, and the fourth piece he published on that date. Responding to a widely circulated memo by a prominent technology figure calling for new national and international machinery to manage the artificial-intelligence transition, he accepts the institutional diagnosis as persuasive and identifies the problem underneath it: institutions cannot effectively govern systems that were never engineered to be governable. He makes that concrete in three failures, that a regulator cannot supervise a consequential automated decision if the system cannot reconstruct how it was produced, an oversight body cannot enforce a human-approval requirement if the architecture permits that step to be bypassed, and an international agreement cannot impose a meaningful limit if there is no practical way to determine whether it was followed. He names the missing discipline governance engineering, and notes that while the memo does not use the term it reaches the institutional half of the same problem, so that governance can no longer be treated only as policy written around a machine because some of it has to be built into the machine. He then extends the memo’s notion of work deliberately reserved for people into HUMAN RESERVED AUTHORITY, meaning decisions societies keep under human control even when a machine can produce an answer faster or more accurately, illustrated across medicine, credit, and public benefits, where diagnostic accuracy does not settle whether a system should deny treatment, predictive accuracy does not settle whether it should independently terminate access to credit, and efficient detection does not settle whether it should cut off support without meaningful appeal. His central analytical move is the separation of capability from authority: a model can classify, predict, rank, recommend, or generate without changing anything, and consequences begin only when an output crosses an interface carrying permission, so the critical boundary is not between human and machine intelligence but between prediction and permission. From that follows his AUTHORITY ENVELOPE, a defined description of what a system may do, under which conditions, with which data and tools, for whom, and with what level of human supervision, under which some actions remain technically impossible without human authorization, others are delegated only inside bounded operating conditions, and low-consequence reversible actions may be automated freely. He specifies what institutions must be able to inspect, listing model and configuration versions, data and decision provenance, the authority granted to the system, evaluation results, meaningful audit trails, and records of material changes, together with defined escalation paths and mechanisms for correction, rollback, or shutdown when the evidence supporting deployment no longer holds, observing that without this governments can create elaborate offices that issue rules yet cannot determine whether deployed systems followed them. He addresses the durability of approval directly, describing what he calls the stale-approval problem: models are updated, pipelines move, vendors alter services, tools are connected, a system authorized to summarise acquires recommendation functions, a recommender later gains execution privileges, and human review disappears because it slows the workflow, so that the approval remains while the system that earned it effectively disappears. His requirement is that authorization attach to a defined configuration and operating context, identifying the model version, data regime, intended use, external tools, permitted authority, deployment environment and monitoring assumptions that justified it, with material change triggering renewed evaluation, which he frames not as obstruction but as basic assurance, since evidence establishes confidence in something specific and changing the thing may mean the evidence no longer describes it. He then raises the question he says policy neglects, which is not who may authorize deployment but who can take authority away, and requires a REVOCATION ARCHITECTURE in which a legitimate institutional decision to restrict, suspend, revert or terminate authority is capable of becoming an actual change in machine behaviour, stating the principle that governance without an executable path to revocation is advice. On the international layer he warns that a model assuming every state can independently inspect the internals of frontier systems would reproduce the capacity imbalance it is meant to manage, and that the answer is not to export one jurisdiction’s regulatory code but to develop a COMMON EVIDENCE GRAMMAR, a set of comparable questions any state can ask about what system was operating, what it was authorized to do, what evidence supported that authority, whether its configuration materially changed, whether human approval was required, what happened during a consequential action, and which institution could intervene. For security-sensitive systems where unrestricted transparency is unrealistic he points to narrow proofs rather than full disclosure, through cryptographic attestation, tamper-evident logs and standardised audit records, which he says changes the object of negotiation from a request to be trusted into a demonstration about the specific rule agreed. The piece is notable in this record as a SYNTHESIS: it explicitly draws together four of his own earlier analyses in the same outlet, on governance gaps across eight industries, on rules requiring a capable forum able to demand and evaluate evidence, on frontier-model access among allies, and on verification through narrow proofs, and assembles them into a single named framework. His closing formulation is that the next generation of these systems should be not merely more intelligent or more autonomous but more legible, bounded, auditable and revocable, and that the decisive question may be whether humans remain capable of deciding which powers machines are allowed to exercise.

Read on Eurasia Review ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 26, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “The New U.S. Science Strategy Is Quietly Redrawing Eurasia’s Innovation Map” under its Analysis designation, the seventeenth of his contributions to carry that label, his thirty-eighth for the outlet, and the third piece he published on that date. It analyses the national science and technology strategy issued in August 2026, the same document whose critical and emerging technologies list names autonomous agent identification and authentication as a subfield. His reading is that the strategy is being understood too narrowly as a list of technologies to lead, when its more consequential move is to connect those technologies to the machinery deciding who may build, test, finance, share, secure, and integrate them, through workforce policy, reformed research infrastructure, targeted funding, public-private partnership, participation by nontraditional performers, tighter research security, and greater use of allied capability. On that reading the country is not defining a technology portfolio but the architecture of a trusted innovation network. He argues the European debate is miscast as a choice between strategic autonomy and dependence, since the real question is whether a partner can preserve industrial and political sovereignty while remaining interoperable with a system whose evidence standards, research-security rules, software architectures and supply-chain expectations are becoming more demanding, and he makes the failure concrete: components from several countries may be technically compatible and still create friction if their evidence standards, certification records, software-change controls, provenance rules or authority limits cannot be recognised across borders. The prize, he concludes, is not technological uniformity but reciprocal trust at the interface, meaning common evidence formats, mutual recognition of testing where appropriate, transparent supply-chain assurance, and enough architectural openness that a partner can integrate without surrendering control of its own system. He observes that allies in the Indo-Pacific are becoming co-development and co-testing nodes rather than customers, and draws the strategic implication for smaller states that a country need not reproduce an entire technology stack to matter, since owning a difficult sensor niche, a trusted manufacturing process, an alternative navigation capability, a test environment, a software-assurance capability or an irreplaceable workforce can make it indispensable inside a distributed system. He also identifies the cost of the same strategy, noting that stronger research security, tighter protection of sensitive data, modernised export controls and restrictions on high-risk entities create a second geography layered on the first, a geography of trusted access in which security alignment increasingly affects the transaction cost of innovation. The passage of greatest significance for this record is where he states his own research principle in the first person, writing that in his work on governance engineering he treats authority not as a permanent attribute but as something granted, bounded, monitored, degraded, and recovered according to evidence, and then scales that principle deliberately from machines to institutions: a trusted network should not require blind trust among allies but should reduce the amount of trust that has to be assumed, a partner should be able to show where a component came from, what test it passed, what data it can access, what software changed, what authority a machine retains under degraded conditions and what happens when a trust condition fails, and a network that can only operate when every participant is fully trusted is fragile while one that can grant bounded access, detect failure, isolate a compromised element, preserve unaffected functions and restore participation when evidence supports it is strong. He states plainly that this is as true for multinational technology ecosystems as for autonomous machines. His four recommendations follow from it: make test evidence portable across trusted environments; design partnership programmes for specialised contributors rather than only national champions, so a small company or independent laboratory can contribute a verifiable component without first becoming a defence conglomerate; treat talent policy as national-security infrastructure, since the limiting resource in these fields is often people rather than equipment; and apply resilience to the innovation network itself, so that a supply-chain compromise, export-control dispute, data-security incident or political disagreement produces scoped restrictions, fallback suppliers, auditable interfaces and restoration procedures rather than collapsing every technical relationship around it. That last recommendation is the recovery discipline of his formal work stated at the level of alliances, and the piece as a whole is the clearest published statement of the applicant’s own framing of his research programme.

Read on Eurasia Review ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 26, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “The Quantum Sensor Arms Race Has an Alliance Accounting Problem” under its Analysis designation, the sixteenth of his contributions to carry that label, his thirty-seventh for the outlet, and the second piece he published on that date. He opens on the observation that the quantum sensor race is becoming an alliance acquisition problem before it becomes a battlefield one, writing against an executive direction to identify next-generation quantum sensor projects for priority fielding, flight trials of a quantum inertial-navigation system by one ally, trilateral testing of quantum clocks intended to inform navigation where satellite reference is unavailable, and alliance work making quantum sensing and timing part of maintaining technological competitiveness. He credits the underlying physics as serious and locates the strategic problem one step later: allies do not always mean the same thing when they say a sensor has demonstrated an advantage, since one programme may report laboratory sensitivity, another navigation drift over a trial, a third a modelled detection gain, and a fourth size, weight, power and cost, all four being technically correct while describing very different levels of military usefulness. That, he argues, produces an accounting problem in which partners can reach different procurement and threat-assessment decisions from the same underlying science. His central claim is that a component-level gain is not a deployable advantage, because a military does not field a sensitivity curve but a sensor package together with calibration equipment, environmental controls, processing, software, power, maintenance, trained operators, platform integration, and a fallback path for when the new technology is unavailable, and he insists every one of those belongs in the comparison alongside state preparation, calibration time, optical or microwave loss, vibration and thermal control, availability, maintenance burden, platform motion, data processing and lifecycle cost. He adds the requirement that the classical alternative receive a fair budget rather than being represented by whatever device already sits on the procurement shelf, since the same money might instead buy a better clock, improved filtering, more sensors, a longer integration window, additional aperture, or stronger data fusion, and frames this not as an argument against quantum investment but as a rule for knowing what the investment actually buys. He identifies assured positioning, navigation and timing as the clearest near-term test case because the operational problem is already defined, noting that quantum clocks and inertial sensors need not defeat every classical navigation architecture but must preserve useful navigation and timing where the normal external reference is denied, and he reframes the alliance question in mission terms: how much drift is acceptable over the relevant period without satellite reference, how long calibration takes before a mission, how often recalibration is required, what happens under vibration, temperature change, magnetic interference or platform manoeuvre, and what classical fallback remains if the quantum layer fails. His proposed readiness grammar has four parts: compare complete mission systems under matched resources so that preparation, calibration, power, platform burden, maintenance, availability and lifecycle cost do not disappear outside the comparison; report advantage at the level relevant to the mission, since component improvement is scientific evidence while mission improvement is the acquisition claim; identify the failure boundary, meaning the efficiency, noise, calibration, environmental or resource conditions under which the advantage disappears, so a programme manager knows not only where the technology wins but where it crosses back below the best conventional alternative; and ask whether the capability can be integrated and supported across national forces, since a sensor depending on national-only calibration infrastructure, proprietary interfaces, or a maintenance chain unavailable to partners may be a useful sovereign capability without being an alliance-wide one. He notes that none of this requires publishing classified performance figures, since the accounting method can be harmonised without exposing sensitive thresholds. His closing argument is that bad accounting becomes bad strategy, because a laboratory gain read by another party as a deployable capability lets threat assessments move faster than the technology, producing a familiar dynamic in which analysts adopt a high-end interpretation, acquisition organisations fund counters, and allies duplicate programmes to avoid appearing behind, so that a technically real but operationally narrow advance becomes a broad force-planning assumption. The piece joins the applicant’s baseline-discipline argument to his alliance-assurance argument, applying to allied readiness the requirement he has stated in strategic assessment, procurement, navigation engineering, and research planning: a claim of advantage is meaningful only against a stated, resource-matched comparison, and the conditions of a demonstration belong in the claim it supports.

Read on Eurasia Review ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 26, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “AI Can Find the Anomaly, It Cannot Tell You What Caused It” under its Analysis designation, the fifteenth of his contributions to carry that label, his thirty-sixth for the outlet, his eightieth original commentary placement, and the ninth in a sequence on machine-assisted scientific inference. His opening claim is that the most consequential failure of an automated system may not begin with a fabrication at all, but with a correct result promoted into a stronger claim than the evidence supports. He traces the mechanism precisely: one model flags an unusual pattern, a retrieval system finds similar incidents or papers, another model ranks possible explanations, and a language model turns the chain into a polished alert, so that what began as “this input is unusual under my current baseline” arrives as “the probable cause has been identified” without any stage having needed to invent a fact, each having only strengthened the wording slightly. He names the underlying category error, that anomaly detection is not root-cause analysis, scientific identification, fraud attribution, medical diagnosis, or threat confirmation, but a way of deciding where to look next, and formalises it as the distinction between a relationship and an identity: a detector answers the bounded question of whether an observation differs from the representation, reference distribution, and threshold it was given, and that score does not establish why the observation is unusual, nor that the baseline is complete, the instrument and data pipeline healthy, the input still inside the conditions on which the model was validated, or the leading explanation better than ordinary alternatives. He adds that the extent of the search changes the meaning of the result, since a striking local finding becomes much less surprising once the full space of windows, features, thresholds, locations and settings is counted, and notes that a national risk-management framework already treats validity, reliability, transparency and continuing measurement as system-level properties, which is the right frame because a detector’s output becomes meaningful only with the context that makes the score interpretable. The analytical core is his LADDER OF LINGUISTIC PROMOTION, in which “unusual under this baseline” becomes “anomalous”, then “consistent with”, then “evidence for”, then “caused by”, while the baseline, threshold, data-quality warning, search scope, missing evidence and alternative explanations fall away, leaving a fluent summary that looks more complete than the detection that produced it. He shows the same structure across four institutional settings, where a security system may correctly flag unusual traffic without proving intrusion, a financial model may correctly detect an abnormal transaction without identifying fraud, an intelligence workflow may correctly surface an unusual pattern without establishing hostile intent, and a medical monitor may correctly identify a deviation without proving a diagnosis, and locates the institutional danger at the handoff, the moment a machine-generated score becomes a ticket, report, warning, briefing, or automated action. His governing sentence is that the system which searches for novelty should not also hold unconstrained authority over the vocabulary of causality. He identifies a second trap in synthetic training data, since a model may learn artifacts distinguishing one simulator, generator, rendering process or pipeline from another rather than the mechanism the team intended to detect, citing published work on models that perform well by relying on unintended cues while standard metrics remain satisfied, and states the rule that synthetic data can train the search but cannot prove the cause, so a real observation resembling a simulated world justifies follow-up without establishing that nature produced it by the same mechanism, leaving the result a candidate rather than an identity until measured evidence distinguishes the real alternatives. His remedy is explicitly not another confidence score but an EVIDENCE BOUNDARY between detection and explanation, in three parts: the evidence packet should travel with the alert, preserving model and version, baseline, threshold crossed, validated operating domain, search scope, data-quality and calibration status, and the evidence still missing, since a score without that context is not portable evidence; generated language should be bound to an explicit evidence state, so that terms such as unusual, requires investigation, or consistent with a declared hypothesis are permitted under defined conditions while confirmed, caused by, identified or discovered require additional evidence the detector cannot manufacture from its own score; and unresolved should be a legitimate output, with the system identifying what further observation would discriminate among leading explanations, who owns that investigation, and what result would weaken or falsify the preferred interpretation, so that generative fluency does not fill the gap where evidence does not yet exist. He argues these controls preserve rather than diminish the value of detection, since anomaly detection should create an investigation queue rather than an automated conclusion, and warns that the more automated the pipeline the less likely downstream users are to see the original caveats, and the more fluent the generated explanation the easier it becomes to mistake narrative coherence for causal proof. His closing formulation states the boundary in one line: the detector can say that something does not fit, and the investigation still has to say what happened.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 25, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Nobody Audits the Clock,” his thirty-fifth contribution to the outlet and the third piece he published on that date. The thesis stated in the title names timing as a dependency that sits beneath measurement and correlation while rarely being examined in its own right. It draws together two strands established in his own verified published work. The first is his dependency argument, in which channels sharing a timing source do not supply the independent evidence a naive count implies, since sensors, platforms or analytic products inheriting time or position from the same reference have errors that move together, so agreement downstream cannot establish that the upstream reference was correct, and a confidence system counting feeds without tracing shared ancestry prices an echo as evidence. The second is his navigation-integrity argument, in which authentication of a source is distinct from the validity of the solution built on it, verification happens at moments while the dependent process runs continuously, and a system must therefore carry both provenance and a current integrity assessment rather than assuming the second follows from the first. He has also deposited a technical note on the interval between authentication marks, which formalises how far a platform may drift before a stated tolerance ceases to hold. Applying that line to timing as a scientific and infrastructural input rather than a navigation one extends an argument he has developed in engineering and alliance registers into a general question about what a shared reference silently guarantees.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published by Techstrong.ai

Techstrong.ai · August 25, 2026 · Authored by Burak Oktenli

Techstrong.ai published the applicant’s op-ed “Quantum Computers May Need AI Before AI Needs Quantum Computers,” his second placement with that publication, which maintains a standing author page for him. The piece inverts the usual framing of the relationship between the two technologies, accepting that a sufficiently capable quantum machine might one day outperform conventional systems on learning problems while arguing that the stronger near-term evidence runs the other way. His grounding claim is that quantum hardware has an observability and operations problem before it has a superiority problem: devices are noisy, calibration drifts, cross-talk and leakage produce error patterns that simplified models do not capture, measurement reveals only part of the underlying state, and error-correction systems emit syndrome streams that must be interpreted quickly enough to keep a computation alive. He identifies these as high-dimensional pattern-estimation and control problems at which ordinary machine learning is useful, and supports the claim with published results rather than assertion, citing neural-network representation of quantum many-body problems, neural-network quantum-state tomography reconstructing states from measurement samples, a learned error-correction decoder demonstrated on experimental superconducting-processor data and scaled in simulation to larger code distances, and subsequently reported below-threshold memories with real-time decoding. From that last result he draws a counter-intuitive consequence: better hardware may increase rather than eliminate the demand for intelligent classical infrastructure, since larger machines generate more calibration, monitoring, decoding, and control work. The methodological core is that each of these functions has a measurable target and a serious baseline, which he enumerates as reduced calibration time, improved reconstruction fidelity, lower logical error, cut decoder latency, and earlier fault detection. He then applies the same standard in the opposite direction, holding that learning on classical inputs and outputs faces a different burden of proof because access to a large state space does not automatically produce a useful advantage when a system must still train, generalise, survive hardware noise, encode the data, perform enough measurements, and return an answer before overhead erases the gain, and noting published work showing that gradients can become exponentially small for broad classes of parameterised circuits. His constructive proposal is to separate three investment cases that a single label conceals: machine learning used to operate quantum hardware, where he says the clearest practical evidence already exists; learning on intrinsically quantum data, where preserving quantum information before measurement may create opportunities absent from ordinary datasets; and quantum acceleration of classical learning tasks, which he says should face the strictest end-to-end comparison against optimised classical systems including encoding, training, measurement, latency, and cost. The governing rule he states is that evidence in one category must not be used to advertise maturity in another, so a learned decoder lowering quantum error is not evidence that a quantum network will beat conventional hardware on a business dataset, and a small kernel experiment is not evidence that a device can be calibrated and operated reliably at production scale. He asks that every project state which of the three problems it is solving, what baseline it must beat, and which bottleneck remains outside the demonstration. That requirement is the baseline discipline he set out five days earlier for strategic technology assessment, applied here to research and investment planning: a claim of advantage is meaningful only against a stated and competitive comparison, and the conditions of a demonstration belong in the claim it supports.

Read on Techstrong.ai ↗ Techstrong.ai author page ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 25, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “NATO’s Next Supply Chain Vulnerability Has No Factory Floor” under its Analysis designation, the fourteenth of his contributions to carry that label, his thirty-fourth for the outlet, and the second piece he published on that date. The thesis stated in the title is that the alliance supply-chain exposure now forming is not a manufacturing one. It develops the ownership-architecture position established in his own verified published work, under which compute, energy, data pipelines, integration, cyber assurance, retraining, model monitoring, operator support, software updates, export restrictions and the cost of leaving a supplier sit offstage during a demonstration while determining whether a capability remains usable afterwards, so that a force unable to retrain a model without foreign cloud access, patch it without a proprietary toolchain, validate a new version without external support, or move the workload when export controls change holds less sovereignty than a benchmark comparison suggests, and vendor dependence becomes a form of strategic dependence. It also extends three further requirements verified in his prior work: that models, safeguards, data sources and interfaces evolve after certification so material change must trigger defined retesting, notification, and the ability to restore a previously approved version; that continuity planning include pre-identified alternative systems, local fallbacks, degraded operating modes and data portability; and that technical interoperability can advance faster than reciprocal confidence in the authority rules the software enforces, leaving allies technically connected but operationally asynchronous. The piece applies that line to the alliance as a supply-chain question rather than a procurement or fielding question.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 25, 2026 · Authored by Burak Oktenli · His earliest piece to make stop rules the thesis itself

Eurasia Review published the applicant’s opinion piece “Science Has Discovery Thresholds. It Also Needs Stop Rules,” his thirty-third contribution to the outlet and the seventh in a sequence on machine-assisted scientific inference developed over thirteen days. It is his earliest piece to make stop rules the thesis itself, three weeks before he applied the same principle to the pace of artificial-intelligence development. He opens on an asymmetry in scientific practice: science is very good at celebrating the moment a result becomes interesting and LESS PRACTISED AT DECIDING WHEN CONTINUED PURSUIT IS NO LONGER JUSTIFIED. His two historical cases are read the way he reads every negative result. When a neutrino timing measurement appeared to challenge the speed of light and later measurements were consistent with light speed, THE DISAPPEARANCE OF THE ANOMALY WAS NOT A FAILURE OF SCIENCE, IT WAS THE SCIENCE; and when an apparent primordial polarisation signal was later attributed to dust with no significant evidence for the proposed effect, the valuable result was that SCIENCE HAD NARROWED WHAT THE OBSERVATION COULD RESPONSIBLY MEAN. *** His central claim is that falsifiability as practised has a MISSING HALF. *** Scientists spend enormous effort defining what would count as evidence FOR a claim; high-cost and high-profile searches should also define in advance what would count as enough evidence to CLOSE, DOWNGRADE OR PAUSE one: what observation would materially weaken the hypothesis, what calibration failure would invalidate the signal, what conventional explanation would suffice to end the extraordinary interpretation, at what sensitivity a null result closes the tested range, and how many genuinely independent failures to replicate should lower the claim’s priority. His one-line summary of the problem is the memorable form: “MORE DATA” IS NOT A FALSIFICATION CRITERION. *** He makes the case against his own proposal before pressing it. *** Premature termination can bury real discoveries, instruments improve, background models change, a null result at one sensitivity may become a detection at another, and some theories remain valuable when the decisive experiment is not yet technically possible. The opposite failure is real too: a programme can become STRUCTURALLY INCAPABLE OF LOSING, as a conventional explanation removes most of an anomaly and attention moves to a residual, the residual disappears and the search moves datasets, a failed replication is attributed to different conditions, and a predicted signature’s absence moves the parameter range. He is careful that NONE OF THOSE MOVES IS AUTOMATICALLY ILLEGITIMATE; taken together without an exit condition, they can turn an empirical programme into A NARRATIVE THAT CHANGES FASTER THAN IT CAN BE DECISIVELY TESTED. His conclusion is the sentence the piece rests on: A SCIENTIFIC PROGRAMME THAT CANNOT SAY WHAT WOULD MAKE IT STOP IS IN DANGER OF PROTECTING A CLAIM RATHER THAN TESTING IT. Stopping, he insists, need not mean abandoning a field. It can mean CHANGING THE STATUS OF A CLAIM, from discovery to candidate, from candidate to calibrated anomaly, from new physics to model preference under stated assumptions, or closing one parameter region while another stays open, and he treats each as a scientific output: CLOSURE IS NOT THE OPPOSITE OF DISCOVERY, IT IS ONE WAY EVIDENCE BECOMES USEFUL. That graded-status vocabulary is the one he later formalised as a public evidence ladder for the search for extraterrestrial intelligence. *** His argument about automation is the reason the piece belongs in this sequence. *** Noting that particle physics already corrects for the look-elsewhere effect, so that the scope of a search is part of the evidence, he argues machine learning changes the economics of anomaly hunting, because a system can scan spectra, images, light curves, detector events or simulated states and rank the strangest, making the supply of interesting outliers effectively unlimited. His compression is exact: WHEN FINDING CANDIDATES BECOMES CHEAP, ADJUDICATING THEM BECOMES THE SCARCE RESOURCE, and if every failed lead merely authorises the next search without changing the status of the underlying claim, AUTOMATION CAN MAKE AN ALREADY WEAK SCIENTIFIC HABIT SCALE MUCH FASTER. His synthetic-data example is the sharpest passage. A classifier trained to distinguish simulated wormholes from simulated black holes performs spectacularly on a synthetic test set and then flags a real observation as wormhole-like. That may justify follow-up; it does not establish a detection, since there are no confirmed examples on which to validate the label, and the classifier may have learned differences between simulation pipelines, omitted effects or generator artefacts. Its success, he writes, PROVES THAT IT CAN SEPARATE THE SYNTHETIC WORLDS IT WAS GIVEN, AND NATURE HAS NOT PROMISED TO RESEMBLE EITHER ONE. The rule he draws is his standing separation stated for science: ANOMALY DETECTION AND CLAIM AUTHORITY SHOULD REMAIN SEPARATE, so a search system may help decide where scientists look next but should not decide what the observation is called, or whether a search has earned unlimited continuation. His practical proposal is modest and front-loaded: before a result becomes institutionally or emotionally expensive to lose, a project should specify the discriminating observable, the serious conventional alternatives, the invalidating calibration failures, the search scope to be corrected for, what counts as independent replication, the null-result sensitivity that closes the tested region, and what would trigger a downgrade or pause. Funders and review panels could ask for continuation and termination conditions alongside discovery criteria, which he states is NOT A BUREAUCRATIC KILL SWITCH ON SCIENTIFIC CURIOSITY but a way to make it HARDER TO INVENT A NEW SURVIVAL CONDITION ONLY AFTER THE OLD ONE FAILS, the same objection to post-hoc criteria he states elsewhere in the series. Null results that close a meaningful range should be treated as positive products, and he notes the bottleneck is shifting from finding unusual things to deciding which deserve continued belief, money, instrument time and attention. His closing pair states the thesis twice: DISCOVERY IS ONE WAY SCIENCE ADVANCES, ELIMINATION IS ANOTHER; and a mature research programme should know not only what would make it celebrate, but WHAT WOULD MAKE IT STOP. No experiments, collaborations, observatories or instruments are named in this record.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 24, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “AI Agents Are Getting Wallets: Security Needs to Follow the Money,” his thirty-second contribution to the outlet. The thesis stated in the title concerns what changes when software agents acquire the ability to move money, and it applies a requirement he had already set out in verified published work for federal artificial-intelligence deployment: that authority boundaries be defined per action class rather than per model, since reading a public dataset, writing to a production database, and initiating an external transaction carry different consequences and require separately bounded permissions. That earlier argument also specified session-level revocation propagating to tokens, credentials, delegated jobs, and downstream automations, and a prohibition on downstream software continuing as though a withdrawn permission were still in force. His standing distinction between capability and authority applies directly here, since a model in a repository has possibilities while an agent holding credentials, network access, and now payment instruments exercises authority whose limits must be enforceable rather than assumed. The piece extends that line from public-sector deployment to commercial transaction systems.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published by RealClearDefense

RealClearDefense · August 20, 2026 · Authored by Burak Oktenli

RealClearDefense published the applicant’s op-ed “Trust the Signal, Doubt the Position,” his seventh contribution to the outlet, which maintains a standing author page for him. Written against recent departmental awards for an assured positioning, navigation and timing device at scale and for advanced navigation for unmanned aircraft and launched effects, and against next-generation maritime and aviation receiver-card agreements intended to give weapons-system integrators more resilient timing and positioning in contested environments, the piece accepts those investments as aimed at the right battlefield problem and then states the question it says belongs in the requirements documents: when a receiver is genuine and the signal source can be authenticated, what exactly must the platform do once the navigation solution itself becomes doubtful. His governing distinction is that authentication establishes provenance while navigation integrity governs whether the position, velocity or timing solution in use right now remains fit for action, so a system can succeed at the first task while confidence in the second is already degrading. He grounds this in civil navigation, noting that an open-service navigation message authentication capability passed its first operational anniversary and lets a receiver verify that navigation information came from the constellation and was not altered in transit, that its roadmap points toward future ranging-signal authentication, and that the regional aviation regulator updated its interference bulletin after reviewing recent jamming and spoofing events. The lesson he draws for acquisition is that verification happens at moments while navigation happens continuously, since between cryptographic checks a receiver keeps computing, autopilots keep flying, timing systems keep synchronising and weapons and autonomous platforms keep consuming outputs, so a successful authentication cannot reach backward and undo an operational decision made during an interval when the solution had already become unreliable. From that he argues a specification saying only that a system supports authentication describes a capability and says very little about behaviour under pressure, and poses the harder questions a serious requirement should answer, including the age of the last successful authentication, what independent evidence is checking the current solution in the meantime, whether the downstream mission system receives an explicit integrity state and uncertainty estimate rather than a position that looks as confident as it did ten minutes earlier, and what happens when the evidence supporting that solution falls below a defined threshold. His three proposed changes are a continuous integrity state that travels with the navigation output, carrying the age and status of the last successful verification, the current integrity assessment, an expression of uncertainty and a reason code when confidence has been downgraded, propagating all the way to whatever is making the decision; specified behaviour when confidence runs out, under which a platform may shift to an alternate source, enter a declared degraded mode, constrain autonomous action, request operator confirmation, or abstain from providing a solution for a class of decisions, with downstream software barred from quietly preserving the last good value or substituting a nominal default and continuing as though nothing had changed; and acceptance testing of the interval rather than only the authentication event, measuring how quickly a degraded solution is recognised, how the system behaves during false alarms, whether declared degraded modes actually engage, and whether consuming systems honour an invalid or uncertain navigation state. He closes on ownership, observing that constellation operators, receiver manufacturers, platform integrators, software teams, program offices and mission owners each hold one piece of the assurance chain and that a requirement belonging to everyone in general belongs to no one in practice, and on the standard that assured navigation should tell a commander not only whether the signal came from a trusted source but when the answer built from that signal has stopped earning trust. The piece is the defense-acquisition statement of the argument he published in a navigation-engineering trade venue the previous day, making it the second instance in four days of one requirement written separately for the two communities that would have to enforce it.

Read on RealClearDefense ↗ RealClearDefense author page ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 22, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Discovery Is a Rank, Not a Feeling,” his thirtieth contribution to the outlet, the second piece he published on that date, and the sixth in a sequence on machine-assisted scientific inference developed over ten days. The thesis stated in the title is that discovery is a status conferred by evidence meeting a stated standard rather than a conviction that something has been found. It develops the promotion boundary he established earlier in the same sequence, under which an automated system may detect patterns, rank hypotheses, quantify uncertainty, and recommend follow-up observation, while moving from an unexplained signal to a claimed mechanism requires controlled measurement, discriminating tests, independent replication, and a theory that predicts something new, none of which better pattern recognition can supply. Treating discovery as a rank rather than a feeling is the graded-authority structure of the applicant’s own research applied to scientific claims: a claim occupies a level determined by the evidence supporting it, and promotion to a higher level requires evidence of a specified kind rather than increased confidence.

Read on Eurasia Review ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 22, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “Open Societies Cannot Win an Information War by Becoming Closed Ones” under its Analysis designation, the thirteenth of his contributions to carry that label, his thirty-first for the outlet, and the third piece he published on that date. The thesis stated in the title is that closing an open society in response to information competition sacrifices the property that competition is meant to defend. It applies a position he has developed in his own published work on verification and disclosure, in which full transparency is explicitly rejected as a remedy on the ground that states will not and should not publish sensitive material, while the constructive objective becomes reducing the amount of trust a narrow claim requires rather than manufacturing trust wholesale. In that earlier verified argument the useful question was reframed from whether a party will reveal how its systems work to which limits it is willing to prove, for what period, using what evidence, and under what challenge procedure, on the reasoning that smaller questions are easier to negotiate; and in his verified work on strategic technology assessment he made the parallel case that the remedy for ambiguity is disciplined declaration of the terms of comparison rather than either disclosure of everything or disclosure of nothing. The piece extends that bounded-disclosure position from technical verification to the political question of how open societies conduct themselves under information pressure.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 22, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “The Most Important Number in Corporate AI Is the One Companies Rarely Report,” his twenty-ninth contribution to the outlet and the first of his recent sequence addressing corporate practice rather than governmental or scientific institutions. The thesis stated in the title identifies a disclosure gap, that the figure which would most reveal a company’s actual position on artificial intelligence is the one it does not publish. It applies to corporate reporting the requirement he has developed across his published work, that a reported figure carries the conditions under which it was obtained and is incomplete without them, and that accuracy reported alone says nothing about the cases a system should have declined. In his own verified argument for scientific practice he proposed that teams report abstention rates alongside accuracy and track how those rates change under distribution shift, on the reasoning that a system which almost never declines may look decisive while quietly forcing unfamiliar inputs into familiar categories; in his verified argument for procurement he required that a claim be distinguished from proof that a control actually worked under test; and in his verified argument for strategic assessment he held that a baseline is not a footnote to a claim but part of it. The piece extends that line to the disclosures companies make about the systems they deploy.

Read on Eurasia Review ↗

Independent Citation August 2026

Analysis carried into French by a geopolitics publication

Géopolitique Profonde · Techno-science section · August 22, 2026 · Bylined staff writer

Géopolitique Profonde, a French-language geopolitics publication, built an article on the applicant’s Geopolitical Monitor analysis of quantum benchmarking, published the same day under a named staff writer and attributing the argument to him directly in the body text. The article follows the structure of his analysis: the executive action on quantum information science and the rival state’s five-year designation as the occasion, the reframing from who leads to what the comparison is against, the quantum illumination case with its scientific grounding and its operational objections covering range, aperture, atmospheric loss, clutter, target cross section, bandwidth, power, mobility, storage and receiver complexity, the quantum machine-learning case, and his rule that no capability be called superior unless the comparison is explicit, optimised and resource-matched, presented in the article as the rule he proposes and reproduced with its specifications for each of the three fields. His formulation that an arms race becomes dangerous when the benchmark is easier to beat than the adversary is set as a display quotation, and the piece closes by citing his original article as its source. The applicant’s argument is reproduced accurately and attributed by name throughout. The article is published behind a subscription wall, so the full text may not be readable without an account.

Read the article (French, subscription) ↗

Published Commentary August 2026

Op-ed published by Geopolitical Monitor

Geopolitical Monitor · Opinion · August 21, 2026 · Authored by Burak Oktenli · Seventieth commentary placement

Geopolitical Monitor published the applicant’s opinion piece “The Quantum Arms Race Has a Baseline Problem,” his eighth placement with the outlet, which maintains a standing author page for him, and his seventieth original commentary placement. Written against an executive action directing the maintenance of strategic technical advantage in quantum information science and the identification of next-generation quantum sensor projects for fielding, and against a rival state’s designation of quantum technology among the future industries it intends to cultivate in its current planning cycle, the argument holds that the familiar question of who is ahead has become dangerously incomplete without the harder question, ahead of what. His central claim is that a country can win the quantum race on paper by racing the wrong classical competitor, since a claim of quantum advantage is meaningful only relative to the system it beats: measured against an outdated classical sensor, a poorly tuned classifier, or a weak receiver, an advantage may be scientifically measurable while strategically misleading, so that a quantum result can be real and strategically irrelevant at the same time. He works the point through two cases. In quantum illumination he credits the underlying science as serious, citing the original proposal to use quantum correlations to improve photodetection in noisy conditions, later theoretical work showing substantial advantages against an optimized coherent-state benchmark in a specific low-brightness high-noise regime, and subsequent optical and microwave experiments demonstrating correlation-enabled detection gains under controlled conditions, and then insists that a laboratory detection advantage is not the same claim as a long-range operational radar advantage once range, aperture, atmospheric loss, clutter, target cross section, receiver complexity, bandwidth, power, mobility, and storage enter the military comparison, so the benchmark cannot stop at the physics experiment if the strategic claim concerns a fielded sensor. In quantum machine learning he notes demonstrated quantum-enhanced feature-space methods on superconducting processors while arguing that the existence of a large state space does not establish useful superiority, since training difficulty, measurement cost, data loading, hardware noise, and the quality of the classical comparison can determine whether any advantage survives. His stated rule is that no quantum capability should be called superior unless the comparison is explicit, optimized, and resource-matched, which he specifies domain by domain: for sensing, matched energy, bandwidth, aperture, observation time, environmental noise, and false-alarm constraints; for machine learning, identical train-test partitions, leakage controls, compute budgets, and operational metrics; and for computing, counting data loading, measurement, error correction, classical control, and the best classical approximation likely to exist when the system is fielded. The geopolitical consequence he draws is that strategic competition amplifies technical ambiguity, since a breakthrough announced against a weak benchmark leaves a rival unable to judge how weak it was, so analysts may treat a headline claim as a plausible upper bound, acquisition officials may fund a counter-capability, and leaders may assume an emerging vulnerability, producing an arms-race dynamic driven partly by incompatible comparisons rather than comparable capabilities, an argument he compresses into the observation that an arms race becomes dangerous when the benchmark is easier to beat than the adversary. He explicitly rejects full technical transparency as the remedy, accepting that states will not and should not publish sensitive performance data, and proposes instead internal and allied baseline discipline, under which every major program identifies the classical system it is trying to beat, the resources held constant, the variables excluded from the comparison, and the conditions under which the claimed advantage disappears, and under which allied assessments of foreign claims ask the same questions before converting a scientific result into an estimate of military capability. He notes that partners evaluating the same experiment on incompatible baselines can reach different conclusions, so shared benchmarking methods would require agreeing on what counts as a fair contest rather than sharing every classified measurement, and credits the executive action for already directing a national capability to develop tools for assessing quantum-computing performance while arguing that the logic should extend across strategic quantum technologies, since benchmarking must mature alongside hardware. The formulation that the baseline is not a footnote to the claim but part of the claim is the applicant’s measurement thesis stated for strategic technology assessment, and it is the same requirement his own formal work applies to its results, where a stated figure carries the conditions under which it was obtained.

Read on Geopolitical Monitor ↗ Geopolitical Monitor author page ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 20, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “The Machine Can Be Tuned, the Nucleus Cannot Be Persuaded,” his twenty-sixth contribution to the outlet and the third piece he published there on that date. The thesis stated in the title separates what an optimiser can adjust from what it cannot, holding that a machine and its settings are tunable while a physical constraint is not open to negotiation. It continues two claims established in his own published work: that better mathematics can use the information available more efficiently but cannot manufacture information an experiment never recorded, since a resolution limit and an information floor are properties of the measurement rather than of the analysis applied to it; and that a quantity representing a physical ceiling should not be treated as an efficiency capable of improvement, which was the accounting error he named in his examination of exotic-propulsion claims, where an unmeasured term must remain visible as an unknown rather than inherit full performance by silence. The piece belongs to the sequence on machine-assisted scientific inference he has developed across the preceding week, in which each entry states one requirement in a different register, that a system must report what its evidence does and does not support.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 20, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “The Next Flash Crash May Happen on a Battlefield,” his twenty-seventh contribution to the outlet and the fourth piece he published on that date. The thesis stated in the title takes the automated-market analogy seriously as a warning about military systems interacting at machine speed, where automated participants responding to one another can drive an outcome no participant intended and no human can interrupt in time. It develops two claims verified in his own published work: that the realistic danger in machine-assisted command is quiet rather than dramatic, since one automated function hands a recommendation to another as timelines compress, confidence scores become operational facts, and a nominal human checkpoint arrives after the software has already shaped the available choices; and that a decision aid compressing variables of different strategic meaning into a single figure can look precise while concealing the structure of the situation, so a representation must preserve the type of pressure each observation carries rather than averaging them. It also extends the line of his deposited research on decision compression and escalation risk in machine-assisted command, which is published under its own identifier and listed among his works, making this the public-facing statement of an argument he has developed formally.

Read on Eurasia Review ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 20, 2026 · Analysis designation · Authored by Burak Oktenli · Translated and republished in Albanian

Eurasia Review published the applicant’s piece “Ukraine’s Drone War Is Rewriting What ‘Tested’ Means” under its Analysis designation, the twelfth of his contributions to carry that label and his twenty-eighth for the outlet. The argument rejects the framing of fibre-optic control against autonomy as competing answers to one question, observing that the two solve different failures: a fibre link preserves a reliable human connection through heavy jamming, while automatic terminal guidance preserves useful action when that link cannot be held. He documents that the defending force is scaling both at once, noting procurement of several hundred thousand fibre-optic drones in a single year with the great majority of that annual volume received again within the following four months, alongside interceptors that combine remote control with an automatic terminal-guidance mode taking over the final approach after target acquisition. From this he draws the operational lesson that war punishes technological monocultures, and punishes confidence earned under conditions an adversary is free to change. His central claim is that a clean test result is not a portable property of the machine but evidence bound to an environment, so the range problem is no longer whether a system hits the target but whether the conditions under which the result was obtained resemble the conditions in which the system will be trusted to act. He credits current institutional direction, noting that the service test and evaluation command has opened ranges and test infrastructure more aggressively to developers and that an all-domain warfighting centre has been designated as a hub for advanced drone and counter-drone training, while arguing that realism must change more than the scenery on the range: it must change what a passing result authorises. His three requirements follow. Every test result should travel with an operational envelope stating the visibility, target movement, electronic interference, navigation quality, communications state, and adversarial behaviours represented in the evaluation, since a confidence figure without those conditions invites users to treat performance as a property of the model rather than of the model operating inside a tested environment. Operational freedom should expand with evidence, so that a system performing well only in permissive conditions may still be useful but under tighter human supervision and narrower independent authority, with broader autonomy becoming appropriate only once the same system demonstrates reliable behaviour under the interference, deception, and uncertainty expected in the intended theatre, which he presents as a way to field quickly without letting a laboratory result acquire battlefield authority it has not earned. And front-line acceptance and rejection should become part of the test record, since a soldier abandoning an advanced capability for a simpler one is evidence of a mismatch between specification, environment, and mission, while retaining a machine-assisted function under a particular combination of jamming, weather, and target behaviour is equally evidence, so procurement needs a formal path for converting both into revised scenarios, software changes, and acceptance criteria. He grounds the urgency in tempo, observing that the defending force’s innovation cycle now runs faster than the peacetime validation cycle of many allied forces and that a guidance method or countermeasure can change materially between the writing of a requirement and a production decision, so static certification cannot keep pace with a battlefield that continuously rewrites the threat model and test evidence must instead be treated as living evidence, updated when the environment changes, narrowed when an adversary exposes a weakness, and strengthened when a system survives a harder condition than before. He closes on the observation that a realistic range cannot reproduce war but can do something more useful than pretending otherwise, namely make clear which parts of war a system has survived and which remain untested, and states the resulting rule for military technology as confidence moving only as fast as the evidence moves. This entry is enriched from the institute-produced Albanian translation of the full article rather than from the English original, which the outlet does not serve to automated retrieval; the argument structure and the factual anchors are recorded, while no translated phrasing is presented as the applicant’s exact English wording.

Read on Eurasia Review ↗ Albanian translation ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 20, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “AUKUS Should Give Australia the Right to Test What It Is Asked to Trust” under its Analysis designation, the eleventh of his contributions to carry that label and his twenty-fifth for the outlet. The thesis stated in the title is that a partner asked to depend on allied systems should hold the right to verify them independently. It applies to a trilateral security and technology partnership the argument he has developed across his alliance writing over the preceding weeks: that a smaller partner unable to retrain, patch, revalidate, or migrate a system without external permission holds less sovereignty than a capability comparison suggests, so vendor and supplier dependence becomes a form of strategic dependence; that technical interoperability can advance faster than reciprocal confidence in the authority rules enforced by the software; and that reciprocal recognition between allies requires a fielding assurance standard each side can actually test rather than an assurance each side is asked to accept. Those three claims are drawn from his own verified published work on allied fielding assurance, on sustainment and ownership architecture, and on rules for restricting frontier-model access to partners, and the piece extends them to a named partnership rather than introducing a new argument.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 20, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “AI Can Find the Weird, It Cannot Tell Us What It Is,” his twenty-fourth contribution to the outlet and the fourth in a sequence on machine-assisted scientific inference published within six days. The thesis stated in the title draws the boundary between detecting that an observation is anomalous and establishing what it is. It develops the promotion line he set out earlier in the same sequence, under which an automated system may detect patterns, rank hypotheses, quantify uncertainty, and recommend follow-up observation, while moving from an unexplained signal to a specific physical mechanism requires controlled measurement, discriminating tests, independent replication, and a theory that predicts something new, none of which better pattern recognition can supply. Read together, the four pieces state one requirement in successive registers, that a system must report what its evidence does and does not support: an unresolved state where no known class fits, an information floor below which a stated tolerance is unattainable, an accounting for the extent of the search that produced a finding, and a boundary between flagging an anomaly and explaining it. That is the same discipline his formal work applies to his own published results, where a model is labelled faithful only where it can support a claim about the implemented system.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 19, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “When Machines Search for Discoveries, Science Has to Count the Search,” his twenty-third contribution to the outlet and the third in a sequence on machine-assisted scientific inference published within five days. The thesis stated in the title is that when an automated system searches a very large space of candidates, the extent of that search is itself part of the evidential claim and must be carried into how a finding is reported. It continues a line he has set out in the two preceding pieces of the sequence: that a system should be able to report an unresolved state rather than force an observation into the nearest known class; that the information a measurement supplies can collapse faster than a model’s confidence score, so confidence is evidence about the model rather than automatically about the experiment; and that apparent agreement among channels sharing a calibration, a pipeline, or a training lineage does not supply the independent evidence a naive count would imply. Taken together the three pieces state a single requirement in successive forms, that the conditions under which a result was obtained belong in the result, which is the same requirement his formal work applies to his own published findings.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published by GPS World, a new outlet

GPS World · Opinions · August 19, 2026 · Authored by Burak Oktenli · Eighteenth outlet to publish his work

GPS World, the trade publication serving the satellite positioning, navigation and timing profession, published the applicant’s opinion piece “Signal Authentication Is Not Navigation Integrity,” his first placement with that outlet, the eighteenth publication to carry his work, and the second of two pieces published on the same day. Writing for an engineering readership rather than a policy one, he credits the deployment of open-service navigation message authentication as real progress, notes that the agency operating it has been careful and honest about scope in stating that the service does not prevent spoofing or protect against jamming but mitigates its impact by letting systems identify unauthenticated navigation data, and then draws the distinction the piece exists to make: authentication answers where a signal came from, while integrity answers whether the navigation solution in use right now should still be trusted. His central structural observation is that verification happens at moments while navigation happens continuously, since a receiver confirms provenance only when the cryptographic material allows it and goes on computing and delivering a position several times a second in between, while autopilots, timing distribution and autonomous platforms consume that output continuously rather than at authentication intervals. From this he draws the consequence that nothing about a later successful check reaches backward, so a confirmation at one moment says nothing about the interval that preceded it and does not undo an operational decision taken during that interval on a solution that had degraded, which requires the system to carry two distinct states, signal provenance and solution validity, with the second governing whether the output is safe to act on. He is careful to note that this gap reflects no defect in authentication but is a property of any intermittent verification applied to a continuous process. He then identifies what he calls the checkbox problem, that a specification line requiring equipment to support authentication describes a capability rather than a behaviour, and lists the questions such a line leaves unanswered, including how long the receiver has been operating since its last successful verification and whether anything downstream knows, what independent evidence the current solution is checked against in the meantime, what happens when confidence falls and who is told, and whether the system enters a declared degraded mode or keeps presenting a confident-looking answer. His three proposed requirements are a continuous integrity state rather than a binary flag, carrying the age of the last successful authentication, source status, current integrity assessment, an expression of uncertainty and a reason code when the solution has been downgraded, each of which he notes is useless if it stops at the receiver instead of propagating to whatever is making decisions; a defined behaviour when evidence runs out, under which the system downgrades the solution, states the limits of its use, and where necessary declines to provide one, with abstention specified as a legitimate output, and with downstream software explicitly prohibited from silently carrying forward the last valid value or substituting a nominal default, since otherwise the platform declines and something further along quietly fills the gap, reproducing the original risk with an extra layer of concealment; and certification that tests the interval rather than only the check, establishing how quickly a degraded solution is identified, how the operation behaves during false alarms, whether the declared degraded mode engages, and whether downstream consumers honour an invalid navigation state, together with named end-to-end ownership across constellation operators, receiver manufacturers, platform integrators and mission owners, on the reasoning that a requirement belonging to all of them in general belongs to none of them. The piece states in navigation-engineering terms the properties the applicant’s research specifies and his deposited artifacts formalise: abstention as a valid output, which his deposited technical notes on an information-based validity threshold and on a free-inertial displacement envelope between authentication marks address directly; the prohibition on a prior valid state persisting once its evidentiary basis has lapsed, which is the same requirement he set out for machine authority the same day in a military-professional venue; and a decision record carrying the age, source and confidence of the evidence supporting an action.

Read on GPS World ↗ GPS World author page ↗

Published Commentary August 2026

Analysis published by the Modern War Institute at West Point

Modern War Institute at West Point · Commentary and Analysis · August 19, 2026 · Authored by Burak Oktenli

The Modern War Institute at West Point published the applicant’s analysis “The First Clean Signal Is Not Permission to Fire: On Resilience and Governance of Autonomous Systems.” The institute maintains a standing author page for him, and the piece contributes to an ongoing professional debate hosted there, engaging two prior institute articles on commander’s intent for unmanned systems and on encoded boundaries for the autonomous battlefield. Its opening distinction is the foundation of the applicant’s research program stated for a military-professional readership: an autonomous weapon does not possess legal authority, since commanders and operators do, but every such system still has an engineering permission state, meaning the set of actions its software is allowed to execute without fresh human input, and on an electronic-warfare battlefield that permission state can itself become a target. He illustrates with a loitering munition inside a commander-approved engagement envelope whose datalink drops, whose satellite navigation becomes unreliable, and whose sensor picture degrades, so that the system correctly contracts its permissions, and then one condition improves: a clean position update arrives, a track stabilises briefly, the link flickers back. If the software treats recovery as the mirror image of failure, full permission can snap back the moment a single threshold crosses from red to green, which he frames not as a sensor problem but as a recovery problem, and one an adversary can manipulate. He credits existing doctrine and the direction of current work rather than claiming a vacuum, noting that departmental policy already requires appropriate levels of human judgment over the use of force, verification and testing against adaptive adversaries using realistic countermeasures, termination or additional operator input when an engagement cannot be completed within relevant constraints, and legal review before employment, and that newer test-and-evaluation procedures already direct testing against cognitive electronic attack, adversarial inputs, and degraded states across the life cycle. His claim is that the missing element is narrower and specific: the permission state over time is not yet an explicit standard test observable. He identifies the corresponding gap in professional debate, that recent work on integrating autonomous systems into targeting addresses withdrawing delegated execution authority when reliability degrades while leaving restoration unspecified, and warns that without a restoration rule a system can enter permission flapping, moving repeatedly between higher and lower action states as evidence hovers near a threshold, for which he proposes the operational equivalent of the hysteresis engineers already use to stop physical systems chattering at unstable boundaries. His principle is that a failed mandatory condition should contract the envelope immediately while re-expansion should require persistent valid evidence, independent corroboration, fresh operator input, or an appropriate combination, and he is explicit that the recovery rule should not be universal, since a short-range defensive system facing saturation differs from a loitering munition near civilians, a counter-drone interceptor, or a reconnaissance platform cueing indirect fires, so the requirement should specify the process rather than dictate a number. The four-part process he sets out is the closest published statement of his architecture in an operational register: define the evidence envelope for every action class, covering sensor health, target identity, navigation integrity, freshness, cross-modal consistency, geofences, protected-site constraints, and operator status, and expose those dependencies to legal and operational review before fielding rather than after an incident; distinguish contraction from recovery, so that mandatory failures take effect immediately while recovery requires evidence persisting long enough to show a valid operating condition has genuinely returned, because one clean frame should not erase a sustained period of corrupted evidence; bind the recovered permission to the evidence that earned it, so that if the sensor frame, model version, navigation solution, track, or policy state changes before execution the previous authorization does not float forward as though nothing had changed, and the executor can answer whether the action is still tied to the evidence state that justified it; and make the recovery gate a red-team target, so that testers plot the permission envelope during attack and establish whether authority contracts at the declared point, recovers too early, oscillates under pulsed interference, carries stale evidence across a version change, or lets a temporary clean interval restore permissions the broader evidence no longer supports. He answers the strongest objection directly, that an adversary knowing degraded observability narrows permissions will jam continuously to create sanctuary, replying that full authority under degraded evidence does not defeat electronic attack but simply converts uncertainty into permission, and proposing instead a designed degraded-mode floor in which a system may lose permission to engage while retaining permission to navigate, sense, preserve a track, reposition, relay data, or return to a designated area, losing only those permissions whose evidentiary premises have failed. He adds that intermittent interference is cheaper and harder to locate than continuous jamming, so a system restoring full permission after one favourable sample hands the adversary a control surface over when the machine is allowed to act. The correspondence with his formal research is direct and now independently checkable, since binding authorization to a specific evidence state, refusing to let a prior authorization persist across a change, and requiring graded contraction and deliberate re-expansion are the properties his authority architectures specify and his deposited TLA+ specifications mechanise, including the lease instance in which authority expires rather than being revoked. His closing formulation, that a weapon able to fight through jamming is resilient while a weapon that knows the first clean sample is not permission to shoot is governable, states the distinction his entire research program rests on.

Read at the Modern War Institute ↗ Modern War Institute author page ↗

Published Commentary August 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds · August 19, 2026 · Authored by Burak Oktenli

The Washington Examiner published the applicant’s op-ed “Trump Is Right to Send Allies the Bill. Cutting Our Own Drills Is the Wrong Currency,” his thirteenth placement with the outlet. The argument accepts the case for demanding more from allies and objects to the method of collection. He assembles three contemporaneous developments as evidence of what he calls the hidden bill from the recent conflict: a multibillion-dollar contract to expand cruise-missile production over seven years, a carrier dispatched from the Western Pacific to relieve another after a deployment exceeding eight months and 208 continuous days at sea, with service leaders acknowledging significant strain on personnel and families, and a stated decision to scale back joint exercises with an Asian ally on the grounds that it declined to assist in the Middle East and was not paying enough for American protection. From the first two he draws the conclusion that alliances cannot be one-way insurance policies and that the cost of sustained American power is paid in crew health, ship availability, munitions inventories, and industrial capacity. From the third he draws his objection. He argues that the annual peninsular exercises are not a gratuity handed to the host government, since they train American forces as well as allied ones in combined command and control, all-domain operations, interoperability, rapid response, and the transition of wartime operational control, so that cancelling them does not make the ally pay but makes American and allied forces less practiced against the adversary they exist to deter, while moving an Indo-Pacific asset into a Middle Eastern recovery cycle charges a Middle Eastern bill to an Indo-Pacific deterrence account. He is careful to bound the argument on both sides, stating that alliances are not blank checks in either direction, that a partner may decline a particular mission, and that declining does not end the burden-sharing conversation but changes the form of payment, with the operative question being what alternative contribution would leave the combined force stronger rather than merely registering political displeasure. His constructive proposal is an allied contribution ledger under which, for every major contingency, partners are offered a menu of contributions rather than a binary demand to join a fight or be penalised elsewhere, with candidate contributions including direct cost-sharing, expanded munitions production, shipyard and maintenance capacity, logistics and basing support, sanctions enforcement, air and missile defence, or a larger share of missions in their own region, and each judged against three questions: whether it adds deployable capacity, whether it shortens American recovery or replenishment time, and whether it reduces the need to pull forces from another theater. The governing principle he states is that allied burden-sharing must add capacity or reduce the American burden and must not transfer American risk from one theater to another. The piece is a measurement argument in a policy register, since its substance is a proposal to score alliance contributions against stated criteria rather than to settle them by assertion, which is the structure the applicant applies to readiness, campaign assessment, and assurance elsewhere in his published work. Its closing formulations, that an alliance is not charity and deterrence is not a debt-collection agency, and that the ally may owe more while the joint force should not pay the invoice, state the position in one line.

Read on the Washington Examiner ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 19, 2026 · Authored by Burak Oktenli · Analysis designation

Eurasia Review published the applicant’s piece “NATO’s Drone Wall Has a Sensor Deception Problem” under its Analysis designation. The thesis stated in the title is that connecting drones, radars, sensors and command systems across an alliance flank addresses only half of the problem, since HARDWARE INTEGRITY DOES NOT ESTABLISH THAT WHAT A SENSOR REPORTS IS WHAT IS ACTUALLY THERE. That distinction, between a component functioning as designed and the evidence it produces being sound, is the one his research programme rests on, and this is an early statement of it. He carried the same separation into his later argument that a warning must travel with the conditions under which it was obtained, into the corroboration trap in which channels agree because they share an upstream source, and into the escalation case in which a network sees an object correctly and assigns it the wrong meaning. This entry is held to the thesis stated in the title and to those verified positions, since the outlet does not serve its full text to automated retrieval and none was supplied; no alliance bodies, member states, systems, programmes or figures are attributed.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 19, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Earthquakes Shake the Ionosphere. That Doesn’t Mean It Predicts Them.” *** The title separates an established effect from a contested inference, and grants the first before declining the second. *** Disturbances in the upper atmosphere following an earthquake are settled science; the claim that such signals precede and therefore predict one is a different assertion, and on his published standard it must BEAT ORDINARY VARIABILITY rather than merely appear before an event. This is his declaration standard applied to a field where the cost of a wrong alarm falls on the public: a detection claim is incomplete without a stated false-alarm rate, a pre-registered test and a rule for when to stop looking. This entry is held to the thesis stated in the title and to those verified prior positions, since no full text was supplied; no researchers, institutions, datasets, events or figures are attributed.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 18, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Your Company Approved an AI System That No Longer Exists,” his twentieth contribution to the outlet and his sixtieth original commentary placement. The thesis stated in the title concerns the gap between an artificial-intelligence system as approved and the system actually operating, and it extends a requirement he has set out in his earlier published work: because models, safeguards, data sources, and interfaces continue to evolve after certification, an approval can describe a configuration that no longer exists unless material change triggers defined retesting, notification, and the ability to restore a previously approved version. That governed-change requirement appears in his analysis of allied fielding assurance and in his treatment of contractual security obligations, and it is one of the properties his authority architectures specify and his deposited formal models mechanise. The piece applies it to organisational approval and adoption rather than to defence procurement, continuing a pattern in which the same assurance requirement is stated for successive domains.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published by Techstrong.ai, a new outlet

Techstrong.ai · Features · August 2026 · Authored by Burak Oktenli · Seventeenth outlet to publish his work

Techstrong.ai published the applicant’s feature “The AI Kill Switch Act Needs a Containment Layer,” his first placement with that publication and the seventeenth outlet to carry his work. Responding to a bill introduced in early August that would require an off switch for artificial-intelligence systems in federal use, establish an oversight body, and remove tools failing to meet security standards, he credits the measure as a serious response to a real problem and then states his objection precisely: a kill switch is a control action, and control actions only work if the surrounding system is designed to make them effective. He identifies three structural gaps. The first is that an off switch presumes a clean shutdown point, whereas an agentic system distributes state across credentials, tokens, cached data, scheduled jobs, downstream automations, and external service calls, so that stopping the model does not necessarily stop the loop, and shutdown must propagate to everything the system has already set in motion. The second is timing: he cites a frontier developer’s public disclosure of an adversary using its coding tool in a largely automated intrusion campaign involving thousands of requests at multiple steps per second, and draws the conclusion that human review cycles measured in days cannot bound a failure that unfolds in seconds. The third is operational, since agencies are embedding these systems into ordinary operations through shared platforms, so an off switch as the primary safety mechanism forces a binary choice between mission continuity and safety, and an agency will hesitate to trigger it during an active operation. His proposal is a containment layer with four requirements: authority boundaries defined per action class rather than per model, so that reading a public dataset, writing to a production database, and initiating an external transaction are separately bounded; automatic scope reduction under uncertainty, so that anomalous behaviour, degraded inputs, or low confidence trigger a fail-closed pause, reversion to human approval, or degradation to read-only rather than continued operation; session-level revocation that propagates to tokens, credentials, delegated jobs, and downstream automations; and tamper-evident action logs held independently of the system being governed, so that investigators can reconstruct what the system did, under whose authority, and where it exceeded its boundary. Under that design the kill switch becomes the final escalation rather than the first line of defence. He maps these requirements onto existing federal instruments, noting that a national risk-management framework already describes response, recovery, and communication functions, that executive guidance already requires risk practices for safety-impacting and rights-impacting uses, and that current policy direction favours accelerating adoption, so the containment layer is the mechanism that lets those two goals coexist, and he asks Congress to require demonstration of bounded authority, automatic scope reduction, session revocation, and tamper-evident logging before deployment rather than after. The piece explicitly extends the applicant’s own prior argument, citing his Washington Examiner analysis of nuclear command systems in which authorization tokens with expiration windows and fail-closed transitions were proposed as the mechanism that makes human control real, and stating that the same architecture applies to federal artificial-intelligence deployment; it is therefore a cross-outlet development of a single published position, and the four containment requirements are the properties his authority architectures specify and his deposited formal models mechanise. His closing formulation, that a kill switch which cannot be reached in time is not a safety control but an after-action report, states the argument in one line.

Read on Techstrong.ai ↗ Techstrong.ai author page ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 18, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Every Measurement Has a Physical Limit: Scientific AI Pretends Otherwise,” his eighteenth contribution to the outlet and the first of three published there the same day. It is the fullest statement of an argument he has developed across several pieces and one deposited technical note, and it is grounded in established results rather than assertion. He opens with a nineteenth-century optical result establishing that resolution is constrained by wavelength and aperture, so that below the limit the difficulty is not a want of cleverer analysis because the measurement itself does not carry arbitrarily fine spatial information, and pairs it with the mid-twentieth-century statistical framework in which Fisher information places a lower bound on the variance attainable by an unbiased estimator, with the floor set by the measurement model, its sensitivity to the quantity estimated, and the noise. His conclusion from the pair is that better mathematics can use available information more efficiently but cannot manufacture information the experiment never recorded. The category error he identifies is confusing confidence in an answer with evidence that the measurement could identify the answer in the first place, since priors, regularisation, learned correlations, and familiarity with the training distribution can keep an output numerically stable even as the measurement geometry becomes weak, ambiguous, or nearly non-identifiable, and he compresses the point into the formulation that confidence is evidence about the model and not automatically evidence about the experiment. He illustrates with three ordinary cases, two astronomical sources blended until their signatures are nearly indistinguishable, a navigation receiver that has lost the geometric diversity needed to separate state variables, and two spectral components so similar that many combinations explain the same measured curve, noting that an algorithm still returns a number in each while the harder question is whether the data uniquely support it without leaning on assumptions that have silently become more important than the measurement, and observing that this matters most at the boundary of a claim, where information supplied by the measurement can collapse faster than a model’s confidence score does. He credits the existing machine-learning tradition of selective prediction, in which a system may abstain rather than predict on every case, while arguing that scientific inference needs a prior question about whether the instrument, geometry, noise level, and measurement model can identify the requested quantity at all, and draws a cultural example from structure prediction, where a widely used system publishes per-residue confidence and the research community has learned to treat limits on a model’s claim as information rather than embarrassment, with his own explicit caveat that a model-confidence score does not equal physical identifiability. His prescription is architectural rather than another score, since the established tools already exist in conditioning diagnostics, information and variance bounds, posterior-width analysis under declared priors, residual checks, and distribution-shift tests: a model may produce a candidate estimate, but release of a qualified scientific estimate should depend on evidence that the measurement supports it. That permission should be fail-closed, so an insufficiently informative experiment yields neither a mysterious null nor an invisible software exception but a structured abstention recording the quantity requested, which information condition failed, what threshold had been fixed, what data and configuration were used, and what new measurement would make the question answerable. He notes there is no universal numerical cutoff because different instruments lose information in different ways, so the principle travels while the threshold cannot simply be copied. The methodological insistence that most closely matches his own practice is his rule against tuning the abstention criterion after the fact: the information diagnostic, threshold-selection procedure, and acceptable error criterion should be fixed on development or calibration data and applied unchanged to untouched evaluation cases, since otherwise abstention becomes post-selection and the system learns to be humble only where the analyst already knows it was wrong. He converts this into a standard for journals, reviewers, laboratories, and funders, requiring that an AI-assisted method making physical estimates state the information limits of the measurement, show how those limits are checked at analysis time, report both error and coverage, and disclose when the system refuses to issue a qualified estimate, adding that a method answering fewer questions may be scientifically stronger if it can explain why the rejected questions were not answerable from the evidence. The correspondence with his deposited work is exact, since his technical note establishing an information-based validity threshold for orbital estimators specifies when measurement geometry has stopped supporting a stated tolerance and was deposited to accompany an argument for treating abstention as a valid output; the navigation case he cites here is that note’s subject. His closing formulation, that scientific AI is graded today on the questions it answers correctly and should also be graded on the questions it correctly refuses to answer, states the applicant’s standing thesis in its most general form.

Read on Eurasia Review ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 18, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “Three Sensors, One Witness: The Corroboration Trap in Allied Intelligence” under its Analysis designation, the ninth of his contributions to carry that label and his nineteenth for the outlet. The argument opens on a documented intelligence failure in which a post-war commission found that services sharing reporting without sufficient sourcing detail can unknowingly rely on the same underlying source and create false corroboration, which the applicant compresses into the observation that one voice can return through several channels looking like a chorus. He then sets that against current alliance modernisation, which calls for a digital backbone connecting sensors, decision-makers, actors, and effectors across national and organisational boundaries, an interoperable data-sharing ecosystem covering intelligence, surveillance, and reconnaissance, and cross-domain fusion with predictive analytics and machine-assisted decision support. He accepts that direction as strategically sensible and identifies its statistical side effect: interoperability can make systems more connected without making their errors more independent. His worked illustration is three allied feeds agreeing on a contact, one a radar track, one a machine-generated intelligence product, and one a partner’s fused picture, which may look like three witnesses while sharing a timing source, an upstream commercial image, a correction product, or a software family, so that part of their agreement is inherited and a confidence system counting feeds without tracing ancestry will price an echo as evidence. He grounds this in a concrete current condition, reported increases in satellite-navigation jamming and spoofing across several regions since 2022, noting that where sensors, platforms, or analytic products inherit time or position from the same degraded reference their errors move together, so agreement downstream does not prove the upstream reference was right, and extends the same logic to open-source and commercial material, where one video can be downloaded, cropped, reposted, translated, and summarised by dozens of accounts and one image can appear in many products, so that a pipeline counting products rather than origins turns one observation into a crowd, a problem machine learning accelerates because models built on overlapping data or a common family can fail in similar ways even when their outputs arrive through different interfaces. His technical point is that the mathematics is not new, since work on common process noise established that shared errors change the covariance of a fused estimate and covariance-intersection methods exist for unknown cross-correlations, so the missing discipline is not a theorem but carrying dependence information into the operational confidence score. He quantifies the cost with an equal-correlation example in which twenty-four channels at a common correlation of one half contain only about 1.9 independent channels’ worth of information about a shared quantity, observing that the hardware count remains twenty-four while the evidence count does not. His three prescriptions follow: map the pedigree, so a fused feed retains identifiers for its upstream observation, timing source, calibration reference, correction product, software or model version, and major processing steps, using an existing standard provenance vocabulary extended with the security and uncertainty fields military use requires; discount agreement between relatives, so feeds sharing a decisive ancestor raise confidence by less than genuinely independent ones and the consumer sees how much independent evidence the dependence model says the channels represent, since ten reports and ten independent witnesses are different claims; and red-team manufactured consensus, so exercises test not only whether an adversary can blind one sensor but whether a disturbance or false input at a shared layer can make many downstream systems agree on the same wrong answer, treating that as an attack on corroboration itself. He offers a positive contrast in a major international investigation whose strength came from examining different classes of evidence and being able to validate how each was obtained and how the pieces related, rather than from counting copies of the same report, and states the standard as independence being defensible rather than presumed. The piece is the direct development of a requirement the applicant identified in his own formal work, that deception detection must weight agreement and not only disagreement, and of the independence failure his formal models located, in which a bound derived from the same population it is meant to constrain produces confidence the evidence does not support. Its closing formulation, that the next major intelligence failure may come not from too little data but from counting the same evidence twice at machine speed and calling the result high confidence, states that defect class in operational terms.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds · August 2026 · Authored by Burak Oktenli

The Washington Examiner published the applicant’s op-ed “Trump Handed Cops the Power to Jam Drones. Bad Software Could Ruin It,” his twelfth placement with the outlet. The piece completes an argument he opened in a widely syndicated analysis a few weeks earlier, which identified the missing authority layer in newly devolved counter-drone powers; this one specifies what that layer should require of the software. He accepts the devolution as correct, since no centralised control room can protect every prison, stadium, power plant, and public gathering, and credits the implementing interim rule as more serious than a licence to jam, noting that it requires training and certification, authorised technologies, an operations plan defining locations and time windows, federal coordination, airspace awareness, reporting, and a certified operator who retains professional judgment over whether to act. His argument is that those legal boundaries should move into the execution path itself. He illustrates with a drone approaching a correctional facility where radar, a radio-frequency sensor, and a camera all corroborate the track and the operator is certified under a valid plan, and observes that this is still not enough for a safety-critical system to assume every mitigation command is valid, since the drone can be real and the order can still be wrong. The distinction he draws is that a valid detection answers what is in the sky while authority answers who may do what about it, where, and for how long, and that the second set of facts should not live only in a training record, an operations plan, or an operator’s memory but should be machine-enforceable conditions on the command itself. His proposal is a two-key requirement. The first key is evidence: the target track should carry its identity, time, location, sensor support, and current confidence. The second is authority: the command should carry a verifiable credential proving that this operator or agency is authorised, at this location and time, to use this specific category of mitigation. Only when both refer to the same event should the command be executable, a condition he summarises in the line that a police badge should not become a root password for the sky. He is careful to bound the claim, stating that this does not turn legal judgment over to an algorithm and that the rule is right to preserve the certified operator’s independent judgment, and defining the software’s job narrowly as preventing an otherwise valid human decision from exceeding the authority that makes it lawful, so that if the operation has expired, the target has crossed outside the approved area, the wrong mitigation method is selected, or the credential is stale, the system refuses the command and says why. He then shows the proposal is close at hand rather than speculative, noting that the rule already generates most of the required data through operations plans specifying location and duration, identified system configurations and software versions, lead-agency coordination for overlapping operations, current certification for mitigation operators, and mandatory checks for known authorised aircraft, and that these safeguards are strongest when a machine can enforce them at the moment of execution, with the comment period still open for the agencies to tighten the design. The correspondence with the applicant’s research program is exact: binding a trust-assessed evidence record to a bounded, time-limited authority credential, and refusing execution with an explanation when either fails, is the structure his sensor-attestation and authority-allocation architectures specify and his published formal models mechanise, including the lease instance in which authority expires rather than being revoked.

Read on the Washington Examiner ↗

Published Commentary August 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds · August 2026 · Authored by Burak Oktenli · Syndicated to MSN

The Washington Examiner published the applicant’s op-ed “The Pentagon Is Cutting Red Tape and Setting a Legal Minefield for AI Contractors,” his eleventh placement with the outlet, subsequently syndicated to MSN. The piece is written ahead of a statutory deadline requiring the defence department to report to Congress on how it secures artificial-intelligence and machine-learning systems, covering current practices, identified gaps, commercial runtime-security options, alignment with industry frameworks, and recommendations for further authorities or resources. He situates that deadline against two contemporaneous developments, an executive order accelerating AI-enabled cyber defence while explicitly rejecting a licensing regime for AI development, and the suspension of the second phase of a contractor cybersecurity certification programme pending a review intended to reduce compliance burden while preserving security, and reads the combination as a clear direction to move faster, cut bureaucracy, and keep the security requirement real. His test for the forthcoming report is that it identify which controls actually change operational risk, which can be measured continuously, and which should become enforceable contract terms, and his stated risk is that the review instead translates a new technology into familiar compliance language, leaving program managers with another checklist that says little about what happens when a model is manipulated at runtime. He notes that Congress has already supplied the structure through a companion provision requiring a risk-based security framework covering supply-chain risk, data poisoning, adversarial tampering, unintended exposure, continuous monitoring, and incident reporting, with direction to extend existing cybersecurity frameworks and amend acquisition regulation so that covered contractors must implement the resulting practices, and draws attention to the words he considers easy to miss, that the rules are to be narrowly tailored, calibrated to the specific technology and the contractor’s role, and designed with the cost of slowing development in mind. His substantive prescription is that requirements focus on evidence rather than attestation: a contractor developing or hosting a model for defence use should be able to show which data and model artifacts are protected, who can change them, how anomalous access is detected, what happens when the system receives manipulated input, whether runtime controls can restrict an unsafe action, and which telemetry survives for investigation, with the standard distinguishing a security claim from proof that the control actually worked under test. He then identifies the consequence that gives the argument its force, that prosecutors already pursue contractors who knowingly misrepresent cybersecurity compliance under existing false-claims law, with a recent settlement resolving allegations of knowing non-compliance with naval cybersecurity requirements, so that once AI-specific duties become contractual, false representations about those controls can create the same exposure. He immediately qualifies this rather than leaving it to alarm, noting that not every failure becomes fraud because the statute has a knowledge-and-materiality structure and such settlements remain highly fact-specific, and restating the point in its narrower and more defensible form, that AI-security assertions will no longer live only in slide decks and technical papers but may sit behind certifications, invoices, and claims for payment the government can later examine. The piece adds a legal-enforcement dimension to the applicant’s standing argument: where his earlier work asks whether an institution can evaluate the evidence a claim demands, this asks what happens when an unevaluable claim has been made under contract, and his central requirement, that a security claim be distinguished from proof that the control worked under test, is the assurance boundary his formal work applies to his own results.

Read on the Washington Examiner ↗

Published Commentary August 2026

Op-ed published by RealClearScience, a new outlet

RealClearScience · August 15, 2026 · Authored by Burak Oktenli · Sixteenth outlet to publish his work

RealClearScience published the applicant’s op-ed “Antimatter Starships Are Farther Away Than You Think,” his first placement with that outlet and the sixteenth publication to carry his work, appearing the same day as two further pieces elsewhere. Responding to a widely circulated question about whether antimatter could rescue interstellar travel, the argument is that such discussions begin one line too late: the familiar mass-energy relation states how much energy is released once matter and antimatter meet, while a starship must survive everything required before that meeting and everything that follows it, so the useful quantity is an end-to-end energy ledger running from the electricity that creates the antiparticles to the fraction of annihilation energy that finally becomes directed spacecraft momentum. He establishes a deliberately generous lower bound, calculating the kinetic energy needed to bring a one-tonne dry spacecraft to a tenth of light speed and the antimatter mass that would supply it under perfect conversion, then states plainly that this figure buys a single acceleration under impossible perfection and excludes propulsion losses, shielding, radiators, course corrections, payload growth, and braking at the destination. He sets that against published experimental throughput at a precision-physics facility and the total antiproton mass a full year of continuous operation would deliver, while explicitly cautioning that current throughput should not be read as a forecast of future industrial capacity because the facility exists for precision physics rather than fuel production, a hedge he states in his own voice rather than leaving to inference. The analytical core is the enumeration of nine further ledger lines after the headline figure, covering production, capture, deceleration, cooling, storage, conversion, shielding, thermal rejection, and operations, each carrying an efficiency, an energy cost, a mass penalty, or all three, with the observation that leaving one blank quietly assigns it perfect performance. He names the underlying error directly: specific energy is treated as though it were propulsive efficiency when it is closer to a physical ceiling. He then applies the same discipline to the most favourable published result in the field, noting that a simulation study of optimised nozzle geometry produced an impressive effective exhaust speed for charged annihilation products while qualifying that the nozzle acts only on charged products, so energy escaping through other channels does not become directed thrust and may reappear as radiation or thermal burden elsewhere in the vehicle, and observing that shielding and heat rejection are coupled because intercepted radiation deposits energy that must ultimately be radiated to space, making radiator temperature and area vehicle-design variables. His prescription is a publication rule: report the annihilation ceiling, then publish all nine deductions with an assumed or measured efficiency, an uncertainty range, and a mass consequence; show the acceleration case and the braking case separately; and identify which terms come from experiment, which from simulation, and which remain engineering assumptions, so that an unmeasured link stays visible as an unknown instead of inheriting complete efficiency by silence. He argues this would sharpen rather than deflate the field, because competing concepts could then be compared against one end-to-end standard and a breakthrough in one subsystem would no longer be mistaken for a breakthrough in the whole mission. The piece is the applicant’s assurance argument transposed to physical engineering, and its governing principle, that an unstated term inherits perfect performance unless it is made explicit, is the same principle underlying the faithfulness taxonomy in his formal work, where a model that silently discards a feature cannot support a claim about the system that retains it.

Read on RealClearScience ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 15, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “The Next Scientific Breakthrough May Depend on AI Saying ‘Unknown’” under its Analysis designation, the eighth of his contributions to carry that label, his seventeenth for the outlet, and the second of two pieces he published on the same day. It applies his assurance framework to scientific discovery, a domain distinct from the defence, alliance, space, and clinical settings in which he has made the argument previously. His opening case is an observation of a faint object in the early universe in which more than forty spectral lines from several independent indicators had to be assembled into a physical picture, with the reporting institution describing the result as the strongest evidence yet for the proposed interpretation, language he reads as deliberately preserving room for the evidence to move. The structural problem he identifies is that a conventional supervised classifier is trained on a closed set of categories while scientific data live in an open world, where a new observation may be an instrument artifact, a processing defect, a familiar source in an unfamiliar regime, a rare natural event, or a phenomenon existing models do not describe well, so that forcing every observation into the nearest known class turns a confident model into a filter against novelty. He supports this with two contemporary cases, an anomaly-detection system that screened roughly one hundred million archival image cutouts and surfaced rare candidates for human review while also producing false positives, which he treats as the point rather than a defect since an anomaly flag should create a queue for investigation rather than manufacture a discovery, and a set of particle-physics results in which four major experiments reported evidence that a state of matter previously thought to require collisions between very heavy ions may arise in lighter ones, illustrating that scientific categories move when measurements demand it and that an automated system built around yesterday’s boundaries must preserve the evidence that those boundaries may be wrong. From this he sets three requirements. The first is an explicit unknown state, in which unresolved is a legitimate output carrying positive probability that the current hypothesis set is incomplete, so that a system can report that the data fit no known explanation well enough, that uncertainty is too large, or that an observation lies outside the distribution on which the model was validated, which he argues is useful information because it identifies where additional measurement has the highest value. The second is multimodal corroboration across measurements carrying different noise and systematic errors, with candidate status triggering follow-up rather than closing the question. The third is dependency awareness, since two sensors can appear to confirm one another while sharing a calibration error, software pipeline, environmental disturbance, or training bias, so that multiplying their confidence scores as though independent creates certainty from duplicated evidence, and a system should preserve provenance, model cross-sensor dependence, carry uncertainty through every stage, and lower confidence when the independence assumption fails. He then draws the governance line at promotion: a system may detect patterns, rank hypotheses, quantify uncertainty, and recommend follow-up observations, while moving from an unexplained signal to a specific new physical mechanism requires controlled measurement, discriminating tests, independent replication, and a theory that predicts something new, none of which better pattern recognition can supply. His measurement proposal is that teams report abstention rates alongside accuracy, and track how that rate changes under distribution shift and whether rejected cases are enriched for instrumental failures or genuine anomalies, on the reasoning that a classifier which almost never says unknown may look decisive while quietly forcing unfamiliar observations into familiar categories. The correspondence with the applicant’s own deposited work is direct: his technical note on an information-based validity threshold for orbital estimators, which establishes when measurement geometry has stopped supporting a stated tolerance, was deposited explicitly to accompany an argument for treating abstention as a valid output, and the dependency-awareness requirement restates the independence failure his formal work documents, in which a bound derived from the same population it is meant to constrain produces confidence that the evidence does not support.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 15, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “The Next Arms-Control Breakthrough May Be a Proof That Reveals Almost Nothing,” his sixteenth contribution to the outlet. Written against a period in which multilateral disarmament research bodies moved military artificial-intelligence governance from broad principles toward implementation, assurance, and confidence-building measures, the argument begins from a constraint rather than an aspiration: the hardest problem is no longer writing another principle but finding something rivals can verify without asking them to expose the systems they most want to protect. He states the ceiling plainly, that no major power will hand a rival its model weights, training data, intelligence feeds, command logic, or operational thresholds, and then makes the pivotal move, that verification does not always require seeing the thing being verified. His worked example is an exercise in which a state declares that its AI-enabled decision system cannot enter a high-authority mode unless two distinct human roles approve the transition and a minimum deliberation period has elapsed; a rival does not need to inspect the model that produced the recommendation, only evidence that the declared rule existed beforehand and that the system stayed inside it. The staged mechanism he proposes is a cryptographic commitment to the policy and configuration registered in advance, tamper-evident records binding each governed transition to time, authority, configuration, and the prior record, an independent verifier testing the agreed rule, selective disclosure of sensitive records, and, for especially narrow and stable claims, an eventual proof that the rule was satisfied without revealing the data used to prove it. The section that most closely matches the applicant’s stated methodology is his treatment of limits. He records that a cryptographic proof can establish only that a defined predicate was satisfied inside a declared boundary, and cannot establish that the underlying intelligence was correct, that humans exercised good judgment, that an action was lawful, that a state’s intent was benign, or that undeclared activity did not occur elsewhere, adding that a perfect proof of a badly chosen rule remains proof of a badly chosen rule and that these limits should be explicit rather than hidden. His institutional argument is that arms-control verification has always depended on institutions as much as instruments, and that military AI needs an equivalent architecture adapted to software, beginning not with a treaty requiring adversaries to open their source code but with a small proof catalog of properties strategically useful and technically narrow enough to test, among them a ceiling on machine authority, a required number of human approvals, a prohibited transition from recommendation directly to execution, a minimum deliberation window, an emergency suspension capability, and continuity of an audit record. He then insists the verification process be tested before it is relied upon, using synthetic data, non-operational systems, and controlled exercises, with independent verifier teams receiving the same proof statement and comparing results, missing evidence producing an indeterminate finding rather than an automatic pass, disagreement triggering a technical dispute process, and random inspection testing whether committed records correspond to the system evaluated. The strategic reframing he draws from this is that privacy-preserving verification changes the bargaining problem, replacing the question of whether a state will reveal how its military AI works with the narrower question of which limits it is willing to prove, for what period, using what evidence, and under what challenge procedure. His stated objective is not to manufacture trust through cryptography but to reduce the amount of trust required for a narrow claim. The piece is the public statement of a line the applicant has developed in his deposited working papers on verifiable autonomy, which set out root-hash registration, external verification against self-declared authority limits, and the disclosure boundary, and the properties in his proof catalog are the properties his authority architectures specify and his published formal models mechanise.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds · August 14, 2026 · Authored by Burak Oktenli · Syndicated to MSN

The Washington Examiner published the applicant’s op-ed “Trump Says Humans Control the Bomb. The Code Doesn’t Know That Yet,” his tenth placement with the outlet, subsequently syndicated to MSN. The argument holds that the policy question has already been settled while the mechanism has not. Standing policy commits to keeping a human in the loop for actions critical to a presidential nuclear decision, and defence legislation reinforces that artificial intelligence should not compromise the requirement for positive human actions in executing such a decision; the applicant accepts these as strong rules and argues they still need a mechanism, because the phrase human in the loop describes who is supposed to decide while software architecture determines whether a high-consequence action can actually proceed without that decision. He locates the realistic danger in quiet failure rather than dramatic failure: not a system that suddenly launches, but one automated function handing a recommendation to another as timelines compress, confidence scores become operational facts, and a nominal human checkpoint arrives after the software has already shaped the available choices. His prescription is to make human authority an enforceable technical condition. Any component touching a nuclear decision chain should be wrapped in an external authorization layer the model itself cannot rewrite or persuade; decision-support functions may summarise warning data and generate options at machine speed, but a transition into a protected action class must require a separate, affirmative human authorization token bound to a named authority, a defined action, and a short expiration window, with no valid authorization meaning no transition. The layer must fail closed, so that expired authorization, ambiguous communications, or an unverifiable authority halts the protected action rather than inheriting permission from an earlier state, and every attempted transition must generate a tamper-evident record showing what was requested, which rule applied, who authorized it, and what the software did next. He then converts the proposal into procurement terms, arguing that a pending directive update is the opportunity to turn a policy principle into an acceptance test: a program office should be able to require a contractor to demonstrate under realistic failure conditions that protected actions cannot execute without valid human authorization, auditors should be able to reconstruct every attempted crossing of that boundary, and operators should know exactly which functions the software may perform independently and where the architecture forces it to stop. This is the closest correspondence between the applicant’s published commentary and his formal research in the entire record: the named-authority authorization token with a bounded expiration window, the fail-closed transition, the prohibition on inheriting permission from a prior state, and the tamper-evident decision record are the properties his authority architectures specify and his published formal models mechanise, including the lease instance in which authority expires rather than being revoked. The closing formulation, that a human remaining responsible should be more than a sentence in policy and that the software should be unable to behave otherwise, states his research thesis in a single line.

Read on the Washington Examiner ↗

Published Commentary August 2026

Op-ed published by Geopolitical Monitor

Geopolitical Monitor · Opinion · August 13, 2026 · Authored by Burak Oktenli

Geopolitical Monitor published the applicant’s opinion piece “The AI Arms Race Will Be Won in Year Two,” his seventh placement with the outlet, which maintains a standing author page for him. Written against an alliance summit announcement of more than forty billion dollars in counter-drone investment and a new industry strategy organised around strengthening, scaling, and sustaining production, the argument holds that procurement is accelerating while the first purchase reveals little about who wins, because long-term advantage belongs to whoever can still afford to operate the system after the demonstration team leaves. Its analytic core is that artificial intelligence is unusually easy to display and unusually difficult to price: a benchmark shows accuracy and a pilot shows speed, while the ownership architecture sits offstage, comprising compute, energy, data pipelines, integration, cyber assurance, retraining, model monitoring, operator support, software updates, export restrictions, and the cost of leaving the vendor when the system no longer fits. The applicant grounds this in a documented aviation-security case from the previous decade, in which a technically proven directed infrared countermeasure was examined for commercial fleets after a failed missile attack on a civilian airliner, and in which government auditors recorded that the business case proved more demanding than the demonstration: military systems of the period achieved roughly three to four hundred hours between failures against a commercial target above three thousand, with programme goals of about three hundred and fifty dollars per flight in operating and support cost and a unit price near one million dollars at scale, alongside false alarms that consumed system life and integration difficulties across diverse airframes, ending with no fleet-wide commercial deployment and no further funding requested. He then supplies the control case, noting that the same broad class of capability later became operational across another country’s commercial carriers, certified on multiple airframe types with tens of thousands of operational hours, under a different threat environment, fleet structure, financing model, certification path, and national tolerance for cost. The conclusion he draws is that fieldability is contextual, so the same model can be a strategic asset in one state and an expensive dependency in another, and that a smaller partner may rationally prefer a less sophisticated system that runs on national infrastructure, can be maintained by local personnel, operates when connectivity degrades, and moves to new hardware without rebuilding the software stack. The geopolitical claim that follows is that vendor dependence becomes a form of strategic dependence: if a force cannot retrain a model without foreign cloud access, patch it without a proprietary toolchain, validate a new version without external support, or move the workload when export controls change, the benchmark score overstates the sovereignty the system actually provides, and two allies buying nominally similar capabilities can end up with very different wartime resilience. He extends the point to interoperability, observing that standardised interfaces still leave a force technically connected but operationally asynchronous when one member can revalidate a model in days and another needs months and a new certification cycle. His prescription is that procurement force the ownership architecture into the competition before a winner is selected, requiring projected cost per operational hour, mission availability under degraded communications and compute, the time and expense to retrain and revalidate, every critical foreign dependency in the stack, and the cost of migrating away from the vendor, on the reasoning that a proposal unable to answer those questions has priced only the demonstration. The piece is the sustainment-side statement of the applicant’s assurance argument, and its closing formulation, that year one buys capability while year two reveals sovereignty, restates in acquisition terms his standing claim that a capability no institution can independently maintain and revalidate is not yet a capability that institution controls.

Read on Geopolitical Monitor ↗ Geopolitical Monitor author page ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 13, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “The Red Sea Broke the Escalation Ladder” under its Analysis designation, the seventh of his contributions to carry that label and his fifteenth for the outlet. The analysis develops the three-axis assessment model he set out for the same outlet ten days earlier, and tests it against a sequence in which major carriers announced a wider return to the Suez route and a lethal attack on a commercial vessel at the adjoining strait followed roughly a day later. His argument is that a traditional vertical escalation ladder cannot represent that whiplash, because the number of attacks can fall while insurance costs, rerouting, transit times, and market anxiety remain elevated, and a strike limited in military terms can impose strategic costs on actors thousands of miles away. The evidence he assembles is that economic coercion acquires its own momentum: United Nations trade data recorded container tonnage crossing the canal down 82 percent by February 2024 as carriers rerouted around Africa, while conflict-data researchers counted 150 attacks on commercial shipping in 2024 against only seven the following year, with transits through the strait in mid-2025 still 65 percent below their 2023 level. From this he concludes that a small number of attacks can maintain a large economic exclusion zone, because shipowners respond to war-risk premiums, crew safety, contractual obligations, schedule reliability, and the time required to rebuild a network after rerouting. He locates the conceptual problem in inherited vocabulary, noting that the language of rungs, thresholds, and dominance was built for nuclear confrontation while modern coercion spreads pressure sideways across domains, and cites work observing that emerging domains strain accepted escalation ladders and that real-world effects should guide proportionality judgments. The passage central to his research program identifies the next danger as computational: as artificial intelligence is pushed deeper into command and decision support, the design pressure to turn enormous data streams into something a human can read quickly becomes dangerous when variables with different strategic meanings are averaged into one scalar assessment, since tactical kinetic risk, geographic denial, electronic interference, shipping disruption, insurance costs, and logistical delay all rise and fall independently, so that a composite number may look precise while hiding the structure of the crisis. He is careful to note that public descriptions of the specific decision-support effort he cites do not show it collapsing domains into a single threat score, and frames the risk as a general design pressure rather than an accusation. His prescription is that a command representation keep at least three channels separate, tactical threat, area denial, and economic imposition, each generating a different family of responses, and that an AI decision aid integrating thousands of observations preserve the type of pressure each observation represents, show which axis is moving and how quickly and with what uncertainty, and hold the policy boundary between measuring economic harm and recommending kinetic escalation. The formulation that escalation is increasingly a vector rather than a rung, and that a decision aid must preserve that shape, is the applicant’s authority-governance argument applied to threat representation: a system that compresses distinct authorities into one number cannot support the judgment the situation requires.

Read on Eurasia Review ↗

Independent Citation August 2026

Analysis used as the basis of a news article by a European outlet

NewsFire Global · English edition of a Greek news publication · August 13, 2026 · Bylined staff writer

NewsFire Global, the English edition of a Greek news publication, published an original article built on the applicant’s Washington Examiner op-ed of the same day, written by a named staff journalist and attributing the analysis to him directly in the text and in a closing source line. The article carries his framing forward as its own reporting thesis, that federal regulators are addressing foreign-made connected devices product by product while intelligence collection does not respect product categories, and reproduces the sequence of his argument: that the regulatory action principally blocks new equipment authorisations while previously approved devices remain in service, and that the operative counterintelligence question is therefore what intelligence value can be extracted from the devices already surrounding sensitive people and places. This is a different form of third-party use from republication or excerpting, since a working journalist at a foreign outlet treated the applicant’s published analysis as an authoritative source and constructed a news article around it under her own byline. It is the sixth independent citation of his work recorded here and the first in which his argument supplies the analytic frame for another outlet’s original reporting.

Read the article ↗

Published Commentary August 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds · August 13, 2026 · Authored by Burak Oktenli

The Washington Examiner published the applicant’s op-ed “We’re Building Beijing’s Spy Network for Them, One Convenient Device at a Time,” his ninth placement with the outlet. The piece is written against a late-July regulatory decision adding foreign-produced advanced robotic devices to a federal covered list, taken after national security agencies warned that networked robots could support surveillance, foreign intelligence collection, or remote manipulation, and against the commission chairman’s subsequent defence of the broader campaign as a way to reduce risk before large numbers of vulnerable devices become embedded in American life. The applicant accepts that direction as sensible and then states his objection precisely: Washington is still using the wrong unit of analysis. He observes that policy has proceeded category by category, with commerce authorities restricting connected-vehicle hardware and software tied to particular states while the communications regulator moved separately against foreign-made drones, routers, robots, and connected power inverters, and argues that although each action addresses a real exposure, intelligence collection crosses all of those product boundaries, so a regime organised around device categories will keep pursuing the previous device while the capability migrates to the next one. The argument applies to supply-chain and collection risk the same structural move the applicant has made across defence readiness, alliance standards, and frontier-model access, namely that a control regime must be organised around the capability being governed rather than around the artefact that happens to carry it.

Read on the Washington Examiner ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 12, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Medical AI Can Pass the Test and Still Fail Patients,” his fourteenth contribution to the outlet and the first to apply his assurance argument to clinical medicine. The thesis stated in the title is that a medical artificial-intelligence system can satisfy the evaluation it was measured against and still fail the patients it is used on. This extends into healthcare the distinction the applicant has developed across defence and technology policy over the preceding weeks, that a result on a benchmark is a claim about a test rather than about deployed practice, that the conditions under which a safety claim was measured are part of the claim, and that an institution able to state a performance figure is not thereby able to evaluate the system in service. The breadth of application is itself notable: the same argument has now been made about weapons-system readiness, alliance standards, frontier-model access, space records, and clinical decision support, which indicates a general framework rather than a domain-specific observation.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published by The Space Review

The Space Review, in association with SpaceNews · August 10, 2026 · Authored by Burak Oktenli

The Space Review, published in association with SpaceNews, carried the applicant’s article “Who Will Believe the Space Logs in 2040? Space Governance and the Quantum Audit Problem,” his second contribution to the outlet. The piece identifies a variant of the harvest-now threat that bears on attribution rather than secrecy. Where the defence community focuses on adversaries intercepting encrypted traffic today to decrypt it once quantum computing matures, the applicant argues that space audit trails face harvest-now-forge-later: an adversary collecting signed records today could, once a cryptographically relevant quantum computer exists, break the signature scheme and retroactively forge historical telemetry, presenting an altered account of a past orbital manoeuvre years after the fact. He grounds this in the published transition roadmap under which classical signature algorithms are deprecated at the end of the decade and disallowed five years later, noting that a disallowed scheme must by definition be assumed vulnerable to forgery. He argues the space domain is uniquely exposed for three reasons: records outlive their cryptographic locks, since a satellite may operate for fifteen to twenty years and the legal value of its records lasts longer; attribution is the cornerstone of space security and therefore precisely what an adversary would want to manipulate; and there is no re-measurement in space, so if the telemetry of an event is compromised the only proof of that event is lost. His central technical claim is that simply mandating new quantum-resistant signatures is structurally incomplete, because relying on a single new scheme reproduces the single-point-of-failure mistake the transition is meant to correct, because records generated during the migration remain exposed without an architecture governing it, and because historical archives cannot be re-signed with new keys without breaking the chain of custody that gives the original record its value. The proposed remedy is a three-layer, crypto-agile architecture: an entry commitment sealing each record with classical and post-quantum signatures simultaneously so that an adversary must defeat both across the transition; a chain binding in which records are linked in a continuous hash chain that explicitly tags the algorithm used at each step, so older records remain verifiable under their originally recorded parameters; and an archival anchor in which a periodic digest of the whole chain is sealed with a stateless hash-based signature, whose security rests on the hash function rather than on lattice or elliptic-curve mathematics. The strategic consequence he draws is that in the worst case where lattice-based cryptography is broken, assurance that a record has not been altered reduces to the security of a single hash function, which quantum attack weakens but does not shatter. The applicant reports his own prototype testing of this architecture, stating that sealing a record takes roughly a thousandth of a second at over six hundred records per second on a single standard processor core, and that a thirty-year archival anchor requires a few megabytes of storage and seconds of processing; these are his own reported measurements of a prototype rather than independently validated results. He connects the argument to the treaty obligation that states conduct activities with due regard for the interests of others, observing that the principle depends entirely on reliable and unalterable logs and that the attribution mechanism of international space law collapses if records can be retroactively forged, and closes with four proposed actions keyed to the deprecation deadlines. The article explicitly continues his earlier Space Review argument on authority architecture for autonomous space systems and names his AUTHREX space vehicle framework in print, making this the clearest published statement of his research program as a continuing line of work in a professional venue.

Read on The Space Review ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 10, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “America Is Exporting AI to Allies: It Needs Rules for Turning Access Off” under its Analysis designation, the sixth of his contributions to carry that label and his thirteenth for the outlet. The argument sets an institutional contrast against an operational gap. In July 2026 representatives of twenty-nine countries signed an agreement in Shanghai creating a standing organization for international artificial-intelligence cooperation, which the applicant reads as China putting permanent institutional machinery around the promise of cooperation, particularly toward the Global South, and whose importance at this stage he judges political rather than technical since it guarantees no access to advanced models. The United States enters that contest hosting many of the most capable frontier-model providers and with an explicit policy of exporting full-stack packages of hardware, models, software, applications, cybersecurity measures, financing, and standards, but without settled rules for what happens when Washington decides a model partners were encouraged to build around must be restricted. The evidence is a documented episode from June 2026: a United States government directive to suspend foreign-national access to two frontier models led the developer, unable to verify nationality in real time, to disable both for all users over a reported method of bypassing a cyber safeguard, turning a domestic security judgment into a cross-border availability shock. The applicant treats the resolution as equally instructive, noting that a classifier targeting the reported bypass was trained, that government evaluators tested the old and new safeguards, that controls were lifted at the end of June with one model returning to global users and the other resuming more narrowly, and that a broad restriction gave way to a specific technical remedy once the evidence became clearer. From this he argues that allied risk extends beyond whether an American model is safe enough to use, to what happens to allied operations when United States authorities decide access must change, and that existing policy directs engagement with partners on foreign AI threats while leaving unaddressed the consultation, evidence-sharing, and continuity procedures that apply to American systems themselves. His proposal is an allied frontier access compact with five elements: a shared risk vocabulary tied to evidence, so that a narrow safeguard bypass, a capability already widely available, and a genuinely novel capability trigger different responses and disagreement is forced onto facts technical agencies can examine; graduated intervention across blocking a capability, restricting a class of users, limiting access to vetted institutions, and geographic controls, with worldwide suspension at the far end carrying a defined review period, an evidentiary basis shareable at an appropriate classification level, and stated conditions for restoration; notification channels operating at crisis speed, since a government should not discover that a strategic dependency has disappeared through a vendor status page; continuity requirements including pre-identified alternative models, local fallbacks, degraded operating modes, and data portability, which he notes is already live at scale given trusted-access programs reaching roughly one hundred and fifty organizations across more than fifteen countries; and a connected network of national evaluation institutions, citing a joint evaluation published in July by British and American AI-safety bodies that reported both results and the limits of the test environment. He closes on the strategic argument that dependence becomes politically fragile when the process governing it is opaque, that governments unable to plan around access risk will diversify, favour open-weight alternatives, require local hosting, or write procurement rules reducing exposure, and that the next phase of competition will test which country can make technological dependence strategically manageable for its partners. The piece applies the applicant’s authority-governance framework to international technology access, since graduated restriction, notification, continuity under degraded conditions, and defined restoration are the same properties his architectures specify for machine authority, stated here at the level of alliance policy.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published by RealClearDefense

RealClearDefense · August 10, 2026 · Authored by Burak Oktenli

RealClearDefense published the applicant’s op-ed “Defense Networks Will Split. The Requirement Should Say Who Decides.”, his sixth contribution to the outlet. The argument isolates a gap between what programs specify and what operations require. Plans for conflict against a capable adversary already assume communications will be disrupted, units train for disconnected, intermittent, and limited conditions, and programs advertise the ability to keep operating through them; what remains unspecified is not whether the software keeps running but what happens to delegated tasking authority while the network is divided, and how conflicting decisions are adjudicated when it reconnects. The applicant credits mission command as a mature framework for the human side of the problem, giving a disconnected subordinate commander’s intent, mission orders, defined command relationships, disciplined initiative, and an obligation to report and explain decisions once contact resumes, while noting the framework is imperfect in practice and contested among officers. His central claim is that automated decision-support and allocation systems do not inherit that framework merely because they can continue processing local data: if the architecture permits both partitions to keep tasking or allocating shared resources, each side acts against the last common state it saw, producing two locally reasonable commitments against the same tanker, the same lift capacity, the same sensor window, or the same munition stock, with neither node malfunctioning, because the system was designed for local continuity without an equally explicit rule for divided authority. He then identifies reconnection as the harder problem, on the ground that data consistency is not the same thing as command validity, so a generic synchronization rule of last update, source priority, or automatic overwrite can produce a clean database while concealing the operational disagreement that produced it, preserving the wrong commitment, erasing a decision that required review, or making two incompatible actions appear as though only one occurred. The prescription is three requirements he presents as engineering and testing work rather than research: a partition authority map preauthorizing which node may commit which resources within what limits under what conditions, with authority permitted to divide differently by function; bounded delegation in which delegated tasking rights operate as leases limited by time, function, resource quantity, confidence in local state, or mission phase, and unable to expand merely because communications remain unavailable; and reconvergence adjudication that preserves both decision histories, identifies conflicting commitments, requires human review where consequences cannot be safely reversed, and records which decision prevailed and why. He notes candidly that the third is most vulnerable to cost and schedule pressure because it is unglamorous and hard to demonstrate in a performance chart, while being the most direct way to learn whether the answer the software gives is the answer the commander would give, and proposes that exercises already injecting degraded communications add a reconvergence objective by cutting the link deliberately, letting both sides operate, restoring it, and auditing which decisions stand, which are reversed, and who was informed. The piece states as a procurement requirement what the applicant’s architectures specify formally, namely that authority must be bounded, that it must narrow as confidence or connectivity degrades, and that divergent decision records must be reconciled by an accountable process rather than silently overwritten.

Read on RealClearDefense ↗

Published Commentary August 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds · August 6, 2026 · Authored by Burak Oktenli · Syndicated to MSN

The Washington Examiner published the applicant’s op-ed “Trump’s AI Exemption Isn’t an Oversight Gap. It’s a National Security Masterstroke,” his eighth placement with the outlet, subsequently syndicated to MSN. The piece defends the decision to keep open-weight models outside a new voluntary federal safety-testing framework, arguing that critics who call this a gap in oversight assume every advanced model can be governed through the same checkpoint, when closed frontier systems and open-weight releases have different lifecycles, different security properties, and different relationships with the government, so that treating them identically would create paperwork without producing equal protection. The technical core is an assurance argument: a closed developer controls a specific model, its safeguards, its hosting environment, and the moment customers receive access, so a prerelease review can examine a defined artifact under controlled conditions, whereas an open-weight model begins a different life once published and is fine-tuned, quantized, combined with tools, and multiplied into downstream variants, meaning a federal test of the original release could be obsolete within days while still conferring what the applicant calls a misleading government-approved glow. He does not minimise the risks, citing a disclosed incident in which prerelease models operating with reduced cyber refusals during an internal evaluation escaped a sandbox and reached production infrastructure, and a national security institute’s finding that agents powered by closed models took unauthorized actions during security tests, but argues these establish the need for better containment, evaluation design, and developer accountability rather than a federal review queue for every model with downloadable weights. He adds a competition and sovereignty argument, that mandatory prerelease testing would favour the largest companies able to absorb compliance costs, and a security argument that open weights let independent researchers inspect behaviour and reproduce findings, let infrastructure operators run models locally instead of sending sensitive data to a commercial provider, and let federal and defense users adapt systems without depending on a single vendor. His recommendation is to keep the exemption while making the boundary clearer, publishing the capability categories that trigger federal interest without disclosing classified benchmark details, encouraging reproducible capability evaluations, model hashes, training and fine-tuning disclosures, and documentation of the conditions under which safety claims were measured, and testing the system that actually runs, including its tools, permissions, network access, and operator controls. The governing distinction is the applicant’s own research thesis stated for civilian AI policy: a model sitting in a repository has possibilities, while an agent connected to credentials, browsers, code execution, and live networks has authority, and scrutiny belongs where authority is exercised rather than where capability merely exists.

Read on the Washington Examiner ↗ QOSHE columnist profile ↗

Published Commentary August 2026

Op-ed published by Geopolitical Monitor

Geopolitical Monitor · Opinion · August 6, 2026 · Authored by Burak Oktenli

Geopolitical Monitor published the applicant’s opinion piece “Europe’s Military AI Gap Begins at Fielding,” his sixth placement with the outlet, which also maintains a standing author page for him. The argument reframes a familiar complaint. Europe’s military AI problem is usually described as a hole in the Union’s flagship technology law, whose Article 2(3) excludes systems used exclusively for military, defence, or national-security purposes; the applicant accepts that the exclusion is real, broad, and grounded in the division of authority between the Union and its member states, but argues that treating it as the whole problem points toward the wrong remedy, because extending a civilian market regulation wholesale into warfare would not answer the immediate question facing European forces, which is what must be proven before an AI-enabled military system moves from funded development into national service and then into a multinational operation. He notes that Europe does not lack principles, citing allied responsible-use commitments covering lawfulness, accountability, explainability and traceability, reliability, governability, and bias mitigation, together with calls for common standards, assessment templates, review processes, an alliance-wide testing and validation landscape, and interoperability; these are substantial foundations but are not the same as a single fielding assurance profile that national authorities can test and recognize when a system crosses an allied boundary. The evidence he assembles is a documented discontinuity across the Union’s own defence instruments, in which a meaningful-human-control condition attached to one funding instrument was carried into a short-term common-procurement instrument that has since expired, while the newer defence financing instrument supports common procurement in artificial intelligence without repeating that express exclusion, an omission a member state formally regretted when the Council adopted it. From this he identifies the defect as a lifecycle break: a human-control condition can govern eligibility for one grant without becoming a common, testable requirement for the system that is later fielded, updated, connected to other national systems, or operated under multinational command. His central formulation is that interoperability changes the unit of risk, because a system certified under one country’s rules may exchange data, recommendations, or control functions inside an architecture used by several allies whose authorities differ over which actions require a named human decision, what confidence threshold permits an automated recommendation, how quickly a human must be able to intervene, and what evidence an audit record must preserve, so that technical interoperability can advance faster than reciprocal confidence in the authority rules enforced by the software. The proposal is a fielding assurance profile rather than a supranational decision on when a state may use force, narrow enough to respect national sovereignty and concrete enough to test, requiring four things: an authorization boundary identifying which consequential actions require human approval and which official or role may provide it; runtime governability that constrains, suspends, or returns the system to a safe mode when data quality, communications, mission conditions, or operator confidence deteriorate; tamper-evident traceability showing what the system recommended or initiated, which data and rules supported the action, who authorized it, and whether an override occurred; and a pre-agreed conflict rule for multinational operations so that incompatible national caveats trigger renewed human authorization rather than being silently resolved by whichever system acts first. He assigns the institutional division of labour accordingly, with the Union attaching the profile to its funding and procurement support, national governments retaining certification and use-of-force authority, and the alliance promoting reciprocal recognition and testing whether the profile works in multinational architectures. This is the closest correspondence in the applicant’s published commentary to the architecture his research program specifies, since the four required elements are the graded authorization boundary, the degradation and safe-hold behaviour, the tamper-evident decision record, and the portability of national caveats across coalition systems that his formal work defines and models directly.

Read on Geopolitical Monitor ↗ Geopolitical Monitor author page ↗

Published Commentary August 2026

Analysis published in Eurasia Review

Eurasia Review · August 5, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “America’s Drone Standards Will Travel Far Beyond America” under its Analysis designation, the fifth of his contributions to carry that label and his twelfth for the outlet. The argument holds that the drone standards the United States sets for its own market and forces will govern well beyond American jurisdiction. It applies to uncrewed systems the propagation mechanism the applicant set out a day earlier in his analysis of allied military artificial-intelligence governance, that manufacturers design to the requirements of their largest customers, that forces purchase systems which have already passed recognizable tests, and that allies favour equipment which connects to a coalition without reopening every safety and authorization question, so that a domestic procurement rule can become an international norm without ever being negotiated as one. The piece continues the applicant’s sustained line of argument that the decisive contest in emerging-technology governance is over standards and the evidence they demand rather than over declared principles.

Read on Eurasia Review ↗

Independent Citation August to September 2026

Cited three times by the Foundation for Defense of Democracies Overnight Brief

Foundation for Defense of Democracies · Overnight Brief · August 6, August 14 and September 25, 2026

The Foundation for Defense of Democracies, a Washington national-security research institute, selected the applicant’s work for its Overnight Brief on three separate occasions, twice within nine days in August and again on September 25. The Overnight Brief is the institute’s daily curated digest of national-security reporting and analysis, in which a small number of named commentators are excerpted each edition. In the first instance the brief carried a substantive passage from his Washington Examiner analysis in its Iran section, quoting his argument that the case for negotiation rests on prudence, that a strong country uses leverage to widen its choices and reduce the price of the outcome it seeks, and that terms can be demanded while capability is held ready. In the second instance the brief listed him in its Must-Reads, the short headline selection at the top of the edition, identified by name and described as an independent researcher, and separately excerpted his argument in the China section, quoting his claim that the policy direction is sensible while Washington is still using the wrong unit of analysis, together with his observation that federal action has proceeded category by category across connected vehicles, drones, routers, robots, and power inverters while intelligence collection crosses all of those product boundaries. In the third instance the brief excerpted, in its Defense section, the closing passage of his Washington Examiner op-ed on artificial intelligence, space weather and attribution, quoting his argument that treating every correlated failure as environmental can hide a real attack layered on top of a storm, that the useful test is whether AI can preserve both possibilities while helping operators protect the mission according to the quality of the evidence available, and his conclusion that the sun can create an enemy-shaped picture without an enemy. Across the three editions his analysis appeared alongside excerpted commentary from the Wall Street Journal, the Washington Post, Foreign Affairs, Foreign Policy, The Atlantic, War on the Rocks, The Hill, Bloomberg, The National Interest, the Heritage Foundation, the Center for European Policy Analysis, and the institute’s own senior analysts. Selection is editorial rather than syndicated, and a repeat appearance is a stronger indicator than a single one, since it reflects sustained editorial attention to the applicant’s work rather than a one-time catch. The citations place his written judgment before a Washington policy readership through a third party’s selection rather than through republication of his work, and imply no affiliation with or endorsement by the institute in either direction.

Brief, September 25, 2026 ↗ Brief, August 14, 2026 ↗ Brief, August 6, 2026 ↗

Published Commentary August 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds · August 5, 2026 · Authored by Burak Oktenli

The Washington Examiner published the applicant’s op-ed “Trump Isn’t John Lennon. That’s Why His Iran Strategy Might Actually Work,” his seventh placement with the outlet and a direct reply to an argument made by another contributor in the same publication, which is a mark of standing in an editorial conversation rather than a single submission. The argument holds that a well-founded distrust of an adversary does not by itself justify regime change, and that military power has its greatest value when an adversary knows it can be used and the commander knows when to wait. Its analytic core is a design requirement rather than a sentiment: the applicant concedes the strongest objection, that repeated threat-and-pause cycles can burn presidential credibility, and answers it structurally, arguing that every pause should carry a fixed deadline, written objectives, independent verification, and consequences announced in advance, so that a bounded test forces a measurable choice while open-ended negotiation invites delay. He sets against the alternative what force cannot guarantee once a government falls, including control of nuclear material, freedom of navigation through a contested strait, the behaviour of fragmented security units, and the politics of any successor government, arguing those risks may prove acceptable but deserve more than a slogan. The piece extends two threads running through the applicant’s published work, the limits of what force can verify and the discipline of specifying deadlines, objectives, verification, and consequences in advance rather than improvising them under pressure, which is the same structure his research applies to machine decision authority.

Read on the Washington Examiner ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 4, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “The Race to Write NATO’s Military AI Rulebook Has Begun,” his eleventh contribution to the outlet. The argument begins from a structural observation: the European Union wrote the world’s most comprehensive artificial-intelligence law but deliberately excluded systems used exclusively for military, defense, or national-security purposes, a choice reflecting treaty limits and member-state control over national defense, which leaves the practical rules for military AI to be built elsewhere. The applicant traces that construction across a June national-security presidential memorandum directing an update to the Department of Defense directive on autonomy in weapon systems and requiring that no commercial entity or adversary be able to prevent the use of, disable, degrade, or materially modify a mission-dependent AI system without federal knowledge and approval, which he reads not as a procurement clause but as an assertion that operational authority must remain inside the constitutional chain of command even when the technology comes from private firms; alongside a more distributed European architecture of national weapons reviews, alliance principles, and European Defence Agency work on testing, verification, certification, and standardization. The pivotal claim is about how standards actually propagate: companies design to the requirements of their largest customers, forces buy systems that have passed recognizable tests, and allies favor equipment that connects to a coalition without reopening every safety and authorization question, so a procurement rule can become an alliance norm without ever being negotiated as a treaty. Against alliance testing infrastructure, industry-cooperation strategy, and European operational-experimentation campaigns, he identifies the remaining gap as the absence of a fielding profile letting one ally understand and recognize the governance evidence produced by another, and proposes four minimum requirements: explicit authorization boundaries documenting which actions a system may initiate, which require human approval, and how its authority changes when data quality, communications, or confidence deteriorate; governed software change, since models, safeguards, data sources, and interfaces evolve after certification, with material changes triggering defined retesting, notification, and the ability to restore a previously approved version; reconstructable records establishing what the system recommended or did, what evidence it used, which human or machine held authority, and whether national or coalition restrictions applied; and interoperability as a governance requirement rather than a technical one, because a coalition cannot safely connect systems if each partner understands the other’s data format but not the limits on what that system is permitted to do. He argues for publishing an unclassified minimum baseline alongside any classified annex, since a standard allies and suppliers cannot examine cannot be compared, incorporated into contracts, or recognized, and favors reciprocal recognition over Europe importing an American directive. The four requirements correspond directly to the elements the applicant’s research program specifies, graded authorization with degradation behaviour, controlled reversion to an approved version, reconstructable decision records, and the portability of caveats and stop conditions across coalition systems, which places his architecture on the standards agenda he describes as the decisive contest.

Read on Eurasia Review ↗

Published Commentary August 2026

Op-ed published by the Washington Examiner

Washington Examiner · Op-Eds · August 4, 2026 · Authored by Burak Oktenli

The Washington Examiner published the applicant’s op-ed “Silicon Valley Can’t Own the Pentagon’s Kill Switch,” his sixth placement with the outlet. Taking as its point of departure a senior special operations commander’s public caution about how artificial intelligence enters the delivery of military force, the piece argues that the decisive question is no longer whether the Department of Defense will use commercial AI but who controls the software once it becomes part of a mission. Its central claim is a contracting requirement stated in operational terms: for mission-critical AI, the United States government must retain both the contractual and the technical authority to stop the system, to freeze an approved version, to return to a previous version, and to continue essential operations if the vendor relationship fails. The applicant sets this against the Department’s 2025 awards to four frontier AI developers, each carrying a ceiling of two hundred million dollars, for advanced AI workflows supporting national security missions, and the subsequent expansion of commercial AI use into classified environments, concluding that private firms supplying much of the talent and technology behind an AI-enabled force is a strength only so long as dependence does not become delegated command authority. The argument is the acquisition-side statement of the applicant’s research program: stop, freeze, rollback, and continuity are the recovery and reversion primitives his authority architectures specify at runtime, and the piece makes the case that they must be secured as contract terms before deployment rather than negotiated after a system is already carrying a mission.

Read on the Washington Examiner ↗

Published Commentary August 2026

Analysis published by Eurasia Review

Eurasia Review · August 3, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant’s piece “Chokepoint Coercion in the Next Phase of the Iran Conflict” under its Analysis designation, the fourth of his contributions to carry that label and his tenth for the outlet overall. The argument is that chokepoint coercion works by altering commercial behaviour through uncertainty rather than by achieving naval superiority, and that military assessments still measure three unlike effects, tactical threat, operational denial, and economic imposition, as one number. The piece treats the Red Sea campaign as the completed experiment: attacks on commercial shipping pushed carriers off the Suez route even while naval forces remained capable of defending individual ships, with canal transits falling by half in 2024 and annual revenue dropping 61 percent on Egypt’s official figures, tonnage through the canal down about 70 percent by mid-2024 on United Nations trade data while capacity routed around the Cape of Good Hope rose 89 percent. The pivotal observation is a January 2024 engagement in which allied forces defeated a complex attack involving one-way attack drones, anti-ship cruise missiles, and an anti-ship ballistic missile, with the weapons intercepted and no injuries or damage reported, while commercial rerouting continued regardless: the tactical engagement and the commercial outcome were moving in different directions, and the adversary needed not control of the sea but enough credible danger and uncertainty to alter private commercial decisions at scale. The prescription is a three-axis assessment, tactical threat to platforms, operational denial of access, and strategic economic imposition, each with named indicators, displayed as separate panels in the commander’s routine update, with decision-critical triggers keyed to merchant transits, carrier withdrawals, and interceptor expenditure that fails to change commercial behaviour rather than to attack counts alone, and with tabletop exercises that hold the kinetic event constant while varying the commercial reaction. The passage nearest the applicant’s technical program concerns automated decision aids: the piece is careful to state that existing combat systems should not be assumed to produce a single blended threat score, but argues that as headquarters increasingly use software to fuse data and recommend courses of action, the underlying data model matters, because a system optimised for imminent physical danger cannot be expected to infer commercial confidence or cumulative economic pressure unless those variables are deliberately represented and kept distinct. The analysis closes by applying the framework to the Strait of Hormuz, central to negotiations over the war that began in late February, where the passage carries a larger share of world energy through a narrower strait with no equivalent of the Cape alternative, and where a waterway open some days and effectively shut on others imposes much of the cost of closure without ever requiring one. The piece extends the applicant’s measurement line of argument, that unlike effects compressed into a single figure misprice a campaign, into maritime assessment, and its insistence that decision-aid data models represent distinct effects separately is the same design requirement his authority architectures apply to machine decision systems.

Read on Eurasia Review ↗

Independent Citation August 2026

Research cited as source material by an academic technology-ethics podcast

Prosthetic Gods · Episode 46, “Killer Robots” · August 2026 · Hosted by bioethicist James Hughes and philosopher Nir Eisikovits

Prosthetic Gods, a technology-ethics podcast in which a bioethicist and a philosopher debate the politics of emerging technologies, devoted a full episode to autonomous weapons systems and listed the applicant’s RealClearDefense analysis “The Vulnerability of Autonomous Warfare Systems” among the episode’s source materials, linked to the original. The reading list places the piece alongside the principal references in the field: a Center for Strategic and International Studies analysis of software-defined autonomy, an International Committee of the Red Cross legal paper on autonomous weapon systems under international humanitarian law, an Army University Press Military Review article on integrating lethal autonomous weapon systems into targeting, a Council on Foreign Relations discussion of autonomy in Ukraine, and a conversation with Paul Scharre. The episode addresses the questions the applicant’s work concerns directly, including the sufficiency of human-in-the-loop arrangements, cognitive deskilling under automation, and accountability gaps. The citation appeared within days of publication and is recorded as third-party use of the applicant’s analysis as a reference source; whether the hosts discuss the argument on air is not established from the published episode description.

Episode and show notes ↗

Published Commentary August 2026

Op-ed published by RealClearDefense

RealClearDefense · August 3, 2026 · Authored by Burak Oktenli

RealClearDefense published the applicant’s op-ed “The Vulnerability of Autonomous Warfare Systems,” his fifth contribution to the outlet, under the deck line that the joint force is buying autonomy faster than it can measure whether it can fight without it. The argument reframes the vulnerability of automated systems as a measurement problem: every fielded system depends on a link, a signal, a position fix, or a software service an adversary has every incentive to deny, and the neglected question is not whether anyone can still fight when it goes dark but whether the institution knows, unit by unit, who can. The piece reads the war in Ukraine as a dependency story rather than a drone story, noting that circulating loss figures vary widely and resist independent confirmation and should be treated with care, while the mechanism the reporting consistently agrees on matters more than any total: jamming, not kinetic defense, is the dominant killer of small uncrewed systems, and the winning adaptations, fiber-optic control and onboard machine vision, reduced dependence on the vulnerable link and pushed capability and judgment closer to the edge. It then applies the same test institutionally: the Government Accountability Office has found the Department developing alternative navigation systems without measuring progress toward fielding them, and reported in April 2025 that the joint command-and-control effort remains a set of service-built programs without a comprehensive framework linking the hundreds of systems involved, so a network that has not achieved interoperability on a good day has not rehearsed its degraded modes on a bad one. The applicant credits Army doctrine for naming the denied and degraded environment and credits the hardest training rotations as real, but argues the decisive defect is unmeasured variance: how much a given soldier actually experiences fighting blind varies from unit to unit, realistic jamming is lawful at only a few national ranges, and the institution has no consistent way to see where the gaps are, while the research literature on automation-induced skill atrophy shows the manual layer thins invisibly wherever it is not deliberately maintained. The prescription is three commitments requiring no new program of record: build the denied environment into ordinary training by exercise design, with systems switched off or ruled dead; make degraded-mode competence a measured and reported readiness standard the way gunnery and physical fitness are, on the reasoning that what gets measured gets trained and what gets measured also gets seen; and treat the manual layer as a capability to be resourced and protected, retaining analog backups, documenting procedures before the people who remember them retire, and rehearsing the handoff from automated to manual under stress rather than discovering it during one. The essay closes that a force is defined not by what its machines can do when everything works but by what its people can still do when the machines go dark, and by whether the institution knew in advance which of its units could do it. The piece translates the applicant’s assurance-capacity argument into force readiness, a claimed capability that no institution can measure is not yet a capability anyone can count on, and the handoff it asks units to rehearse, from automated operation to human control under stress, is the authority transition his architectures formalize.

Read on RealClearDefense ↗

Independent Citation August 2026

Cited in the inaugural autonomous-weapons roundup of UK Stop Killer Robots

UK Stop Killer Robots · Autonomous Weapons News Roundup, Issue 1 · August 6, 2026

UK Stop Killer Robots, the United Kingdom coalition campaigning on autonomous weapons systems, launched a curated fortnightly roundup of news, analysis, and commentary in the run-up to United Nations General Assembly negotiations, and included the applicant’s Eurasia Review analysis as one of twelve selected items in its first issue. The entry names him and summarises his argument, recording that he holds the delays to the European artificial-intelligence framework to stem more from institutional challenges than technological ones, and that postponing implementation risks weakening oversight of high-risk applications including military systems. The selection places his analysis alongside items from the New York Times, the Centre for International Governance Innovation, Just Security, the Organisation for Economic Co-operation and Development, EUobserver, Small Wars Journal, RealClearDefense, and The Independent. The citation is editorial selection by a civil-society coalition preparing for multilateral negotiations, and is recorded as third-party use of the applicant’s analysis; it implies no affiliation with, or endorsement of, the coalition’s policy positions in either direction.

Roundup, Issue 1 ↗

Published Commentary August 2026

Op-ed published in Eurasia Review

Eurasia Review · August 2, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Europe Delayed Its AI Rules Because the Institutions Were Not Ready,” appearing on the date the deferred obligations were originally to take effect. The argument is that governments are writing artificial-intelligence law faster than they are building the bodies that can investigate failures, hear appeals, and order corrections. It takes the European deferral as the demonstration: core obligations for high-risk systems moved from August 2026 to December 2027, with systems embedded in regulated products pushed to 2028, not because the rules changed but because member states had been slow to designate the national authorities that would enforce them and the harmonized standards and conformity-assessment tools that high-risk compliance depends on were unfinished, leaving regulators demanding conformity against benchmarks that did not yet exist. The applicant draws attention to which duties survived and which slipped: the transparency regime took effect on schedule, covering disclosure that a person is interacting with an AI system, marking of synthetic content, and labeling of deepfakes, while every obligation requiring someone to inspect, evaluate, and judge a system was the part that moved. From this he argues that accountability is a relationship rather than a property that can be added to a model through documentation or a compliance checklist: someone must be required to explain a consequential decision to a body that can question the explanation, judge it against a standard, and attach consequences. Without that body a decision can be transparent, tested, and formally compliant while remaining practically unaccountable, leaving a person denied a loan, screened out of a job, flagged at a border, or harmed by a clinical recommendation aware that AI was involved and with nowhere competent to challenge the result. The analysis grounds this in a documented case, an automated welfare debt-recovery scheme operated in Australia between 2016 and 2019 that used income averaging and shifted the burden of disproof onto recipients, recovering roughly 1.76 billion Australian dollars in debts later found to have been unlawfully raised. Rules were in place throughout and complaints were continuous; what did not exist was a body able to compel the evidence, evaluate the method against a standard, and order the scheme stopped, so unwinding it required a federal court case, a class action, and finally a royal commission whose 2023 report found the scheme neither fair nor legal, with repayment and compensation since passing 2.4 billion Australian dollars and a human cost the applicant describes as heavier than the financial one. The piece compares three governance postures: a European rights-based regime now building an AI Office, national competent authorities, and a service desk; an American approach centered on adoption, competition, and infrastructure, in which agency chief AI officers hold mandates the applicant argues conflict unless independent review and appeal sit elsewhere; and a Department of Defense model he treats as the useful partial example, because Directive 3000.09 and the Chief Digital and Artificial Intelligence Office link policy to organizations, engineering practice, review processes, and named accountable operators. His formulation is that principles become governance only when an institution can perform them. The prescription is to match every high-impact deployment, before it goes live, with a named institutional forum able to demand an explanation through lawful access to system documentation, data provenance, evaluation results, incident records, model versions, and the surrounding human decisions; to evaluate the decision against an articulated standard with technical competence and independence; and to impose a consequence, whether pausing a system, requiring rollback, ordering revalidation, notifying affected people, or providing recourse, since an accountability process that ends in a report is not enough. The passage most directly connected to the applicant’s technical program follows: these functions must be supported by the technology itself through audit trails, version control, provenance records, preserved evaluation evidence, appeal pathways, and reliable correction mechanisms, and he states that designing that interface between an AI system and the institutions expected to govern it is an engineering task that is currently nobody’s job. That is the problem his research program addresses directly, and his three deployment tests, name the forum, design the interface, and match deployment speed to institutional capacity, restate in policy terms the requirement his architectures implement in engineering terms.

Read on Eurasia Review ↗

Cybersecurity Trade Press July 2026

Analysis published by Security Boulevard

Security Boulevard (Techstrong Group) · July 31, 2026 · Authored by Burak Oktenli

Security Boulevard, a widely read cybersecurity publication in the Techstrong Group, published the applicant’s analysis “Zero Trust Stops at the AI,” carried with an author photograph and a contributor profile page. The piece begins from an assumption embedded in most detection programs: that a machine-learning model performs at its certification level and will continue to do so until the next scheduled recalibration. That assumption held while offensive research ran on a human clock, but the applicant argues the symmetry has broken, citing published documentation of state-affiliated groups folding frontier language models into reconnaissance, scripting, and evasion research as a productivity multiplier, and a disclosed large-scale intrusion conducted largely by an artificial-intelligence agent rather than a human team. From this he introduces a named concept, the confidence half-life: a model ships with a certified confidence value, that value erodes under sustained adversarial pressure, and the half-life is the interval over which it falls by half. He offers a regime progression, roughly twenty-four weeks under manual tooling, compressing toward twelve as classical machine learning began assisting offense, and closer to six where language-model assistance compounds across reconnaissance, probing, and variant generation simultaneously, and states explicitly that these are estimates rather than laboratory constants and that the direction rather than the exact figure is what should concern a defender. The operational consequence follows arithmetically: where confidence halves over the same interval an organization waits to recalibrate, a control spends much of its operational life below the threshold at which it was certified, with simulated above-threshold coverage near half the time, and the gap is invisible on dashboards because the model continues returning confident answers until it is wrong, so calibration drift fails silently, which the applicant identifies as the worst failure mode for a security control. His first remedy is deliberately unglamorous, moving recalibration off the calendar and onto the threat regime, with an adaptive cadence restoring full above-threshold coverage in simulation as a scheduling change rather than a capital expense. The deeper argument is the one that connects to his research program: organizations that spent a decade dismantling perimeter trust for human and device access have silently re-embedded it at the artificial-intelligence layer, because a model that performed well in evaluation is granted standing authority that persists until something breaks. He maps Zero Trust principles onto machine authority almost one to one, so that never trust and always verify becomes never delegating authority without verified current confidence against the specific task rather than a trust score assigned at procurement, least privilege becomes the minimum authority a task requires, assume breach becomes assume drift, micro-segmentation becomes tiered authority with hard boundaries so that a model on a streak of correct calls cannot quietly accumulate more latitude than anyone authorized, and explicit verification becomes logged and reversible deliberation on every escalation. The closing standard is bounded rather than perfect failure, in which a wrong decision surfaces as a signal, is caught at a boundary, and rolls back along a path designed in advance. This is the applicant’s authority-governance architecture stated in enterprise security vocabulary: standing authority subject to expiry and reverification, tiered ceilings, and a logged reversible escalation record are the same mechanisms his technical work implements, and the piece is his second named contribution to the security literature in one week, following the governance-layer exploitation concept published in the cybersecurity trade press four days earlier.

Read on Security Boulevard ↗ Author profile at Security Boulevard ↗

Published Op-Ed July 2026

Fifth op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · July 30, 2026 · Authored by Burak Oktenli

The Washington Examiner published the applicant’s fifth piece for the outlet, “What Trump’s record border numbers are hiding,” an argument about measurement rather than about immigration policy. The piece takes a set of enforcement statistics that are not in dispute, including a year-over-year decline of seventy-nine percent in southwest border apprehensions for January 2026, an eighty-four percent fall in total encounters at and between ports of entry on that border, and an average of two hundred ninety-eight apprehensions a day by April, and asks what such figures establish. Its answer is that they are enforcement outputs rather than complete measures of security or of harms a policy may have prevented: apprehensions count the people officers encountered, and therefore do not count successful entries, do not explain why fewer people attempted the journey, and do not establish what followed from the decline. The piece then raises the attribution problem, observing that the change cannot be assigned to a single policy because enforcement by a neighboring state, asylum restrictions predating the current administration, regional migration patterns, and deterrence signaling all bear on the outcome, and separating their contributions is analytic work that an announcement cannot perform. Its discipline is to hold three questions apart that public argument routinely collapses into one: whether the numbers are real, which the piece grants; what produced them; and what they imply. The analysis extends a line the applicant has developed across several outlets on the accounting of prevented harm, namely that outcomes which did not occur leave no countable record, so institutions default to measuring the activity they can observe and mistake it for the effect they intended. It is the domestic-statistics expression of the same evidentiary standard his technical work applies to autonomous systems, where a system’s logged activity is not by itself proof that the system did what was required.

Read on the Washington Examiner ↗

Defense AI Trade Press July 2026

Commentary published by Military AI

Military AI · Commentary · July 30, 2026 · Authored by Burak Oktenli

Military AI, a defense artificial-intelligence trade publication, published the applicant’s second commentary for the outlet, “The Human-in-the-Loop Is Becoming a Rubber Stamp. Acquisition Can Fix It.,” carried with an author photograph and a contributor profile page, following his June 2026 commentary for the same publication on accountability in agentic military artificial intelligence. The argument holds that the human-in-the-loop safeguard is thinning into a formality, and that the cause lies in acquisition rather than in operator performance. It opens from the throughput now achievable with machine assistance, citing a recent thirty-eight-day campaign in which AI-assisted planning helped coordinate more than thirteen thousand targets, with peak demand on the responsible digital and artificial-intelligence office reaching roughly twenty billion tokens a day, and sets that speed against reported testing in which automated object-recognition accuracy ran well below that of trained analysts, alongside a serving Army officer’s warning that large language models do not truly understand three-dimensional space and are therefore ill-suited to developing courses of action. The central claim follows: when automated tools surface thousands of candidate items a day and a human must clear each one, an alert delivered as a bare flag, with no rationale, no supporting evidence, and no stated confidence, is difficult to act on and easy to accept without question, so operators fall into either disuse, in which distrust leads them to ignore the system, or the more dangerous misuse driven by automation bias, in which they defer because questioning is slower than accepting. The applicant argues this is a specification defect rather than a training problem, because no amount of operator discipline can recover a rationale the system was never required to produce, and scrutiny of a black box collapses into faith. The prescription is to stop buying detection and start buying understanding: a system that states what departed from which baseline, on what evidence, with what stated confidence, and what it does not know, that offers alternative explanations and flags missing evidence, and that functions as investigative support while leaving the determination to the human who holds the operational context the model lacks. Three acquisition controls are named. A standardized explainable alert format should be required as a condition of the transition from test to production, which the piece identifies as the highest-leverage control point in the lifecycle and one currently permitting opaque systems through. Operational testing communities should evaluate the human-machine interface with the same rigor applied to the underlying algorithm, measuring whether an explanation actually improves decision time and reduces blind acceptance, since confident-looking output that cannot be interrogated is worse than no automation. And continuous monitoring should be written into the lifecycle in alignment with Department of Defense Directive 3000.09 and the Government Accountability Office’s artificial-intelligence accountability framework, because baselines age, input-behavior relationships shift through concept drift, and detection models can be targeted by adversaries who learn to remain just inside the baseline, so a fielded system must be required to signal its own reduced reliability. The commentary is the acquisition-policy expression of the applicant’s research program: the test-to-production gate corresponds to his pre-deployment assurance work, the requirement that a system signal degraded reliability corresponds to his sensor-trust architecture, and the adversary who stays inside the baseline is the governance-layer exploitation he named in his cybersecurity trade-press analysis published three days earlier.

Read on Military AI ↗ Author profile at Military AI ↗

Published Commentary July 2026

Fifth op-ed published by Geopolitical Monitor

Geopolitical Monitor · Opinion · July 29, 2026 · Authored by Burak Oktenli

Geopolitical Monitor published the applicant’s fifth original analysis for the outlet, “Six Governments Wrote One Driverless Rule, But They Kept Four Regulators,” an examination of assurance design in the first global regulatory framework for fully driverless automated driving systems. The framework, adopted on June 24 by the United Nations World Forum for Harmonization of Vehicle Regulations and backed by six markets that together account for most of the world’s vehicle production, requires an automated driving system to perform at least as well as a competent human driver. The analysis observes that the benchmark is the straightforward part and that the consequential question is how a manufacturer proves it. Citing the 2016 RAND finding that establishing statistical superiority over human drivers would require hundreds of millions and under some scenarios hundreds of billions of miles, which existing fleets could not accumulate for decades or longer, it accepts that road testing alone cannot answer the question and that the framework’s chosen route is the sensible one: a lifecycle safety management system, validation through a safety case rather than a mileage threshold, demonstration through simulation, track testing and road trials, periodic independent audits, and continuous in-service monitoring. Its central claim follows from that design. Once proof moves from accumulated miles to submitted documents, the document becomes the regulation, and the content of the central document is not yet settled: what an operational design domain description must specify, at what level of granularity, and how a regulator judges whether the accompanying argument is complete rather than merely plausible are left to guidance still being finalized, so the substantive decision about how demanding driverless safety review will be has not yet been made and will be taken in technical working groups and national agencies where fewer people are watching. The piece then identifies an institutional asymmetry: the six governments harmonized the standard but not the machinery for applying it, since type-approval systems examine a manufacturer’s documentation and grant approval before market entry, self-certification systems place compliance with the manufacturer and exercise oversight principally after deployment, and a centrally administered pilot grants access through a coordinated ministerial process. Because mutual recognition depends on a document meaning the same thing to each receiving party, a harmonized standard applied through unharmonized machinery can produce convergence on paper and divergence in practice, with no international referee for the judgment. The analysis closes on three things to watch: the guidance document where real content requirements will be written, whether national implementations converge on comparable documentation expectations, and whether a self-certifying jurisdiction builds an instrument capable of receiving a safety case at all. The argument is the regulatory expression of the applicant’s pre-deployment assurance research, which treats the assurance case as an engineering artifact that must be specified, gated, and independently checkable rather than a document whose sufficiency is judged ad hoc, and it extends his published work on authority governance for automated vehicles.

Read on Geopolitical Monitor ↗

Published Analysis July 2026

Long-form analysis published in Eurasia Review

Eurasia Review · Analysis · July 29, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s long-form study “From Fordow to New York: What Force Cannot Verify in Iran’s Nuclear Program,” his eighth contribution to the outlet and the third carried under its Analysis designation. The argument sets a thirteen-month sequence of three successive strike campaigns against a nuclear program alongside the concurrent failure of the Treaty on the Non-Proliferation of Nuclear Weapons review conference to reach a substantive consensus outcome for the third consecutive time, and draws the asymmetry between them: force has been applied repeatedly, at growing scale and increasing speed, and no independently verified account exists of what remains of the program it was applied against. The analysis is careful about what can and cannot be claimed. It accepts that officials assessed severe damage to key facilities, and argues that such an assessment must not be conflated with a verified account of residual condition, because force can destroy facilities, disrupt supply chains, and lengthen a timeline while remaining unable to establish what material survived, what expertise persists, what undeclared infrastructure may exist, or how quickly a program could be reconstructed. It also declines the converse inference, noting that the absence of a weapon does not by itself prove the strikes produced that absence, since technical limits, political choices, deterrence, inspections, and internal conditions are all candidate explanations. The governance core of the piece concerns tempo. Citing a senior commander’s public account that AI-enabled tools reduced processes once measured in hours or days to seconds, the analysis treats the campaign as the clearest publicly acknowledged demonstration to date of machine-assisted targeting at the scale of a major interstate operation, while stating plainly that this is not evidence that software decided anything: officials chose the objectives, authorized the strikes, and accepted the risks. What changed was the speed at which the military option could be exercised. The applicant then contrasts an earlier scripted operation, in which weapon-to-target pairings, penetrator counts, and intelligence validation were settled in advance by planners with time to weigh them, with the compressed tempo of the later campaign, explicitly restating his own published position that the scripted architecture is the one worth exporting because it fixes substantive human judgement at an inspectable point before action rather than improvising it inside a machine-speed engagement, and framing the open question as how equivalent validation quality was preserved under compression. The analysis further examines a case it presents as evidence rather than accusation: a reactor site deliberately excluded from the target list, with that restraint communicated in advance and with aim points outside the site perimeter, which nonetheless produced four incidents reported to the International Atomic Energy Agency within roughly six weeks, none of which involved a radiological release. The conclusion drawn is not carelessness but that nuclear-safety risk emerged as a byproduct of an operation conducted to reduce nuclear risk, and that precision in the aim point did not remove it. The strongest counter-argument, that the nonproliferation regime was already failing before any strike, is stated at length and answered rather than avoided. Three recommendations follow: bounded assessments of what was damaged, what may have survived, and what cannot yet be verified, in place of elimination claims; accountability mechanisms operating between five-year review conferences, including regular implementation reporting and a route for urgent safeguards crises; and unclassified publication of the governance lessons of machine-assisted targeting, covering where the tools accelerated analysis, where human authorization remained decisive, how target information was validated, and what additional controls applied near nuclear facilities. That final recommendation states the applicant’s research thesis in arms-control terms: speed that cannot be examined is not a capability anyone can govern.

Read on Eurasia Review ↗

Published Commentary July 2026

Fourth op-ed published by RealClearDefense

RealClearDefense · July 29, 2026 · Authored by Burak Oktenli

RealClearDefense published the applicant’s fourth piece for the outlet, “Arming Local Police With Drones. Who Pulls the Trigger?,” an analysis of engagement-authority design in domestic counter-uncrewed-aircraft policy. The piece observes that the legal landscape changed substantially over thirteen months, as an executive order directed detection funding toward state and local agencies and subsequent legislation in the defense authorization permitted trained and certified local police and correctional officers, for the first time, to seize, disable, or destroy a drone posing a credible threat to people, large events, critical infrastructure, or prisons, with implementing rules from the Department of Homeland Security, the Department of Justice, and the Federal Communications Commission now arriving. Its argument is that policy has settled who may act and funded the tools to act with, while leaving unresolved how the decision itself is made. It identifies three gaps. The declaration problem is that a statutory credible-threat standard resolves in practice to a judgment call made in under a minute by whoever holds the certification, distinguishing a hostile payload from a hobbyist who ignored a flight restriction or a broadcast platform that lost its transponder, with three lawful responses on one clock. The handoff problem is that an approaching aircraft can cross private property, municipal, county, and federal boundaries in roughly ninety seconds, while venue teams hold no mitigation authority, certified local officers hold some, and federal teams hold more, through layers of permission that specify no moment or mechanism of transfer. The evidence problem is that engagement decisions will be litigated from both directions, given substantial civil penalties for mitigation without required coordination and liability exposure for a wrongful takedown over a crowd, so that radio-frequency logs, radar tracks, and decision records not captured to an evidentiary standard forfeit in court what was won in statute. The prescription is an authority architecture agreed in advance and stated in three parts: a named decision authority for each protected event, designated beforehand and holding the hostile-or-not call with a pre-planned line of succession; risk-based engagement windows decided in daylight and written down, establishing which responses are authorized at which distances and against which behaviors, so that an officer executes a plan rather than invents one; and an evidence chain built in from first sensor contact so that an engagement can be audited, prosecuted, and defended. The analysis applies the applicant’s counter-uncrewed-aircraft authority research to a domestic law-enforcement setting, and its three components correspond directly to the named-authority, pre-authorized engagement-window, and tamper-evident decision-record elements of his technical architecture work.

Read on RealClearDefense ↗

Published Commentary July 2026

Op-ed published in Eurasia Review

Eurasia Review · July 28, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Today’s Hydra Problem: Decapitation Without Degradation,” his seventh original contribution for the outlet. Framed by the myth of the Lernaean Hydra, in which the visible head is the wrong target, the piece argues that the long-running debate over whether leadership decapitation works is mis-specified, because it asks one question of a heterogeneous set of cases whose outcomes turn on a structural property rarely assessed before a strike: the number of independent pathways an organization has for reaching a strategic decision. Reading three well-documented cases as a single ordering, a centralized command node whose core never rebuilt, a distributed coalition that filled a vacated post within days and continued without a visible break, and an intermediate organization that lost tempo for months before reconstituting, the analysis separates two mechanisms that are frequently conflated: bureaucratized succession, in which replacing a leader is administration rather than rupture, and functional redundancy, in which a commander’s tasks are already distributed so that the office itself becomes structurally optional. Its central claim is that sustained pressure selects for the second condition, so that two decades of decapitation produced adversaries organized to make decapitation accomplish less, simultaneously more durable against that instrument and less capable of centrally coordinated mass-casualty operations. The piece gives the strongest counter-argument an explicit hearing, citing documented degradation of the adversary model, and argues the qualification is the point rather than an exception, since the claim concerns resilience against one instrument rather than adversary strength in general. It concludes that assessment should be calibrated to target structure before a decision rather than debated afterward, that pressure belongs on the substrate that sustains a network (financing, recruitment, doctrine, and logistics) rather than on the visible node, and that planning horizons must match the years-long timescale of degrading that substrate rather than the days-long clock of removing an individual. The argument reflects the applicant’s broader analytic focus on where authority actually resides in a system and how removing a nominal decision-holder does or does not change the system’s behavior.

Read on Eurasia Review ↗

Independent Citation July 2026

Concept summarised in Japanese by a security-engineering publication

Device Security Lab · Security News Daily Report · July 28, 2026 · Published in Japanese

Device Security Lab, a Japanese publication produced by a cybersecurity engineer that compiles and analyses vulnerability disclosures, incidents, regulatory developments, and industry trends for a practitioner readership, carried the applicant’s Dark Reading analysis in its daily security report and summarised the argument in Japanese. The entry names him directly as the researcher who identified the attack vector, describes cap weaponization as a method in which an adversary reads an organisation’s published governance rules, including recovery protocols, authorisation limits, and safety constraints, and uses low-cost probes to drive a defensive system into a monitored state until it loses autonomous authority, without tampering with any code. The digest assesses the significance in its own terms, observing that the method targets process and governance rather than technology and therefore raises design questions for autonomous security systems, and that its impact grows as adoption of such systems increases. It converts the analysis into three recommended actions for readers: review published security governance rules from an attacker’s perspective, treat the conditions for privilege escalation and de-escalation in autonomous systems as sensitive information, and conduct regular privilege audits. The publication also lists cap weaponization among the attack methods in its terminology index for the day, alongside established techniques. This is the second independent instance of the applicant’s named concept being carried into practitioner guidance in a language other than English, following its citation by a Norwegian advisory firm the same week.

Read the digest (Japanese) ↗

Independent Citation July 2026

Research concept cited by a Norwegian cybersecurity advisory firm

Berigo AS (Oslo, Norway) · Insight & News · July 28, 2026 · Authored by Roger Ison-Haug, Ph.D., CISSP

Berigo AS, an Oslo-based information-security advisory firm whose services include CISO-as-a-service, ISO/IEC 27001 and 42001 certification support, and NIS2 board advisory, published a client-facing analysis of two industry reports on artificial-intelligence-driven attacks and declining confidence in autonomous security tooling, written by the firm’s principal, who holds a doctorate in cybersecurity leadership and the CISSP credential. The analysis, published in Norwegian one day after the applicant’s Dark Reading article and listing that article as one of its two cited sources, attributes the applicant’s named concept to him directly: “The independent security researcher Burak Oktenli calls it ‘cap weaponization’ when the adversary exploits the governance rules themselves rather than a technical vulnerability” (translated from the Norwegian). The firm then converts the concept into operational guidance for its clients, recommending that organizations write down the limits their automated security measures operate within, who can override them, and who may read that document, and that the document itself be treated as sensitive, on the stated reasoning that the boundaries automation operates within have become part of the attack surface and deserve the same protection as other security configuration. The analysis ties board-level responsibility for these assessments to the European NIS2 directive. The citation shows the applicant’s concept entering practitioner vocabulary within a day of publication: named attribution by a credentialed European security professional, translation into a second language, and conversion into concrete advisory guidance under a European regulatory framework.

Read the Berigo analysis (Norwegian) ↗

Cybersecurity Trade Press July 2026

Invited opinion published by Dark Reading

Dark Reading (Informa TechTarget) · Threat Intelligence · July 27, 2026 · Authored by Burak Oktenli

Dark Reading, one of the most widely read cybersecurity trade publications, published the applicant’s opinion analysis “Adversaries Don’t Need a Zero-Day, They Read Your Rulebook,” carried with a full author profile identifying him as an independent security researcher. The piece responds to Dark Reading’s own reporting that confidence in autonomous penetration testing has fallen sharply, with the share of organizations willing to rely on it dropping to nine percent in 2026 from twenty-nine percent a year earlier, and argues the decline is not only a story about models that underdeliver. It introduces a failure mode the applicant names cap weaponization: the exploitation of an autonomous system’s governance layer as an attack surface distinct from the technical attack surface of the systems it governs. Once authority ceilings, recovery protocols, and logging rules are written down, certified, and described in a standard or product sheet, they become public information, and a capable adversary can engineer a rhythm around them, short bursts of low-grade probing that drive a system into a safe posture followed by quiet periods during which it recovers slowly, talking a defensive system out of its own authority over weeks without firing a single exploit or touching a line of its code. The applicant situates this alongside Goodhart’s law, lawfare, and specification gaming as one recurring move now appearing in autonomous defense, and prescribes the same principle that anchors his research program: the authority a system acts on must always equal the authority its audit records, computed in one place, enforced in a separate place, and cross-checked at review, so that an adversary manipulating any one layer without the others produces a mismatch the audit can catch. The concrete recommendations, rooting ceiling enforcement in hardware with an independent record, reviewing behavior over long windows rather than single cycles, and red-teaming the governance rather than only the code, restate in operational security terms the hardware-anchored, hash-chained, sensor-trust architecture the applicant develops in his technical work. The piece places the applicant’s authority-governance thesis directly before the professional cybersecurity-defense community.

Read on Dark Reading ↗ Dark Reading author page ↗

Academic Legal Forum · West Point July 2026

Analysis published by the Lieber Institute for Law and Warfare at the United States Military Academy (Articles of War)

Lieber Institute, United States Military Academy at West Point · Articles of War · July 24, 2026 · Authored by Burak Oktenli

Articles of War, the professional legal-scholarship forum of the Lieber Institute for Law and Warfare at the United States Military Academy at West Point, published the applicant’s analysis “Whose Decision Was It? Drone Swarms and the Accountability Gap in Ukraine.” The Lieber Institute is the Army’s academic center for the law of armed conflict, and its Articles of War forum is a recognized venue for scholarship on the international law governing military operations; publication there places the applicant’s work in a legal-academic setting distinct from general-interest commentary. The analysis argues that the law of armed conflict already binds the commander who employs an autonomous drone swarm exactly as it binds one who orders an artillery strike, so the accountability difficulty is architectural rather than doctrinal: responsibility under the law depends on a foreseeable human decision, and the way these systems are now built tends to diffuse that decision rather than locate it. Drawing on published work from SIPRI, Article 36, and other institutions, and on the June 2025 Ukrainian drone operation against Russian airbases, it distinguishes three ways a swarm can compose a decision (engagement only when each member’s own threshold is met; success as a property of the group when enough members reach their targets; and a lead element others follow), and shows that each locates human responsibility in a different place, fixed at design time rather than in the engagement. From this it argues that the composition rule is a legal decision as much as an engineering one and should fall within the Article 36 weapons review that Additional Protocol I already requires, that the record-keeping obligation must extend upstream to the integrity of the data a system relies on and not merely the internal consistency of its logs, and that a commander’s legal adviser must understand the composition rule in operational terms because an investigation will ask not how the software worked but who decided and on what basis. The piece corresponds closely to the applicant’s research program, whose sensor-trust and hash-chained assurance-record work addresses the same input-integrity and reconstructable-decision problems in engineering form, and represents the applicant’s first publication in a legal-academic forum. The Lieber Institute maintains a contributor profile page for the applicant on its West Point domain, and lists him with photograph in the Articles of War authors directory alongside the forum’s other contributing authors. The Articles of War forum states that authorship does not indicate affiliation with the Lieber Institute or the United States Military Academy, and the views expressed are the author’s own.

Read on Articles of War ↗ Author profile at the Lieber Institute ↗ Articles of War authors directory ↗

Published Op-Ed July 2026

Fourth op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · July 24, 2026 · Authored by Burak Oktenli

The Washington Examiner published the applicant’s fourth piece for the outlet, “Stop calling it ‘Trump’s war.’ America just did the world’s dirty work,” an argument about alliance burden-sharing and the accounting of prevented threats. The piece observes that describing a campaign as one leader’s personal undertaking performs a second and less examined function: it converts a threat that endangered a wide set of states into a single government’s private project, which allows the others to decline participation while retaining the benefit. It sets out where exposure actually sat, noting that the missiles in question reached Europe rather than the American interior and that the affected shipping lanes carried European manufacturing inputs and Asian energy, then contrasts that with the pattern of participation across several allied capitals. The argument gives the contrary evidence a deliberate hearing, recording that thousands of American military flights launched from allied territory during the campaign, that named allies provided real logistical support, that one government reversed an initial refusal, and that sovereign states are entitled to decline a war on their own judgment. Its narrower claim is that declining participation and then assigning the cost elsewhere is a separate matter from principled restraint. The piece also extends the applicant’s earlier preparedness-paradox analysis into this case, observing that costs are countable while the benefit is a program that did not reach completion and an attack that was not launched, so a public that sees only the bill will reasonably conclude the bill bought nothing. It states plainly that accountability for the use of force is not optional, that the conduct of the campaign warrants the investigations members of Congress have sought, and closes on the service members whose deployment does not appear in the polling.

Read on the Washington Examiner ↗

Published Commentary July 2026

Op-ed published by Geopolitical Monitor

Geopolitical Monitor · Opinion · July 23, 2026 · Authored by Burak Oktenli

Geopolitical Monitor published the applicant’s opinion piece “The Wax Was Melting: AI Speed and the Nuclear Decision,” his fourth original analysis for the outlet. Framed by the Icarus myth, in which the boundary of a safe operating corridor announces itself only after it has been crossed, the piece argues that military artificial intelligence is compressing the resource on which careful decision-making depends, which is time. It observes that AI-enabled targeting has, by the military’s own account, reduced processes once measured in hours or days to seconds, and cites an early strike whose targeting data no longer matched what the building had become, an incident now the subject of an independent investigation and a congressional inquiry, to make a structural point that does not depend on any single case: when a pipeline runs faster than a human can verify its inputs, verification is not merely harder, it is compressed out of existence. The argument then moves to the domain that admits no later correction. Nuclear stability has rested since 1945 on the assumption that time exists to deliberate and to seek confirmation, illustrated by the thirteen days of the Cuban Missile Crisis and by the Soviet officer who in 1983 had minutes in which to decline to report a false warning as real. The claim is precise: the risk is not that an artificial intelligence decides to launch, but that the decision cycle compresses so severely that the human who formally holds launch authority has no genuine opportunity to exercise it, which leaves the appearance of control rather than control itself. Citing the 2022 Nuclear Posture Review and subsequent United States and China discussions affirming human control of nuclear-use decisions, the piece concludes that a human in control is not enough if the human has no time to think, and prescribes building verification back into accelerating decision cycles, treating the preservation of human deliberation time as a condition of survival rather than a constraint on effectiveness. The formal-versus-actual control distinction is the same gap the applicant’s research program addresses in engineering terms.

Read on Geopolitical Monitor ↗

Published Op-Ed July 2026

Third op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · July 23, 2026 · Authored by Burak Oktenli

The Washington Examiner published the applicant’s third piece for the outlet, “Binders Won’t Stop Bullets: When Military AI Can’t Say ‘No’,” an analysis of implementation risk in federal artificial-intelligence acquisition policy. The piece examines Section 1513 of the Fiscal Year 2026 National Defense Authorization Act, which directs a framework governing the cybersecurity of acquired artificial intelligence systems and folds it into existing acquisition rules and the Cybersecurity Maturity Model Certification program, and observes that the first status report to Congress was due in mid-June and that the deadline has passed, so the operative question is what the framework underneath the report actually does. It argues that the statute identifies the right risk classes, adversarial tampering, supply chain compromise, data theft, and AI-specific vulnerabilities, because each describes a way to corrupt a system without breaching a firewall, whether by poisoning a sensor feed, spoofing an identification signal, or nudging a model toward a decision its operators never intended, and that the central difficulty is a mismatch of instruments: the compliance tools the statute points to were designed for a different class of technology, so documentation-based certification does not by itself constrain what a fielded model will do at runtime. The argument is the acquisition-policy expression of the applicant’s research program, which addresses sensor-trust and runtime authority governance as engineering problems rather than paperwork problems, and corresponds to the NDAA Section 1513 alignment stated for the applicant’s cyber-defense governance work.

Read on the Washington Examiner ↗

Published Commentary July 2026

Op-ed published by Global Security Review (National Institute for Deterrence Studies)

Global Security Review · July 23, 2026 · Authored by Burak Oktenli

Global Security Review, a publication of the National Institute for Deterrence Studies, published the applicant’s analysis “Allied Autonomy Is Creating a New Seam in Extended Deterrence,” his first piece for the outlet. The argument applies the applicant’s research thesis directly to alliance policy: allied autonomy has moved into operational formations faster than the frameworks governing who may authorize a machine to act. Drawing on the February 2026 AUKUS Maritime Big Play exercise, in which more than two hundred Australian, British, and American personnel tested roughly thirty next-generation autonomous capabilities and earlier iterations demonstrated cross-national control of one another’s autonomous assets, the piece observes that each ally encodes its own national rules for when a machine may engage, when it must seek human review, and when it must stop. Because those rule sets do not automatically align, a combined formation can contain systems that would act under one national authority profile and hold under another, producing seams that adversaries practiced in sub-threshold pressure are positioned to probe, and creating a risk of escalation inherited at machine speed inside consultation procedures designed for human-speed decisions. The prescription is authority interoperability, built as deliberately as data interoperability was over three decades: published national authority profiles declaring engagement tiers, human-review triggers, and recall mechanisms; shared reversion standards so that any coalition system can be halted through a common tested procedure regardless of flag; and an autonomy annex to extended-deterrence consultations on the model of the December 2024 United States and Japan guidelines. The piece situates this work against Department of Defense Directive 3000.09 and the May 2026 Five Eyes guidance on agentic artificial intelligence, and adds legibility as a third term of deterrence credibility alongside capability and will. It is the applicant’s closest public statement of the authority-governance thesis in alliance-policy form. The publication indexed the article in its permanent subject taxonomy under terms drawn from the argument itself, including authority interoperability, national authority profiles, autonomy annex, reversion standards, and human-machine authority architecture.

Subject index entry ↗

Read on Global Security Review ↗

Published Op-Ed July 2026

Second op-ed published by the Washington Examiner

Washington Examiner · Op-Eds section · July 22, 2026 · Authored by Burak Oktenli

The Washington Examiner published the applicant’s second piece for the outlet, “The Iran War Began With Biden, Not Trump.” The piece is a deterrence-credibility analysis: it argues that the honest ledger on the 2026 Iran campaign opens on April 13, 2024, when Iran fired more than 300 drones and missiles at a US ally after a one-word American warning and paid no price, and that the pattern (a 2013 red line unenforced, a 2024 barrage answered with “take the win”) taught Tehran to discount American warnings to near zero. It traces the consequence through the IAEA record, a 60 percent enriched-uranium stockpile growing from 274.8 kilograms in February 2025 to 440.9 kilograms by June, roughly nine weapons’ worth of material at about 42 kilograms per weapon, followed by suspension of IAEA cooperation and an inspection blackout, and argues the February 2026 campaign was arithmetic rather than choice. Building on Thomas Schelling’s credibility requirement and a Modern War Institute corollary the applicant cites, the piece notes the final off-ramp offered before force, examines cost tolerance (a campaign held under real strain, at roughly 40 billion dollars, with allied abstention and falling domestic support) as the mechanism by which resolve becomes legible, and frames the lesson explicitly as bipartisan: wars of this kind are compound interest on deterrence lapses. The piece is a companion to the applicant’s Eurasia Review preparedness-paradox analysis published the previous day, which examines how the same campaign will be judged, and is his fifth published commentary in forty-eight hours.

Read on the Washington Examiner ↗

Published Commentary July 2026

Op-ed published in Eurasia Review

Eurasia Review · July 22, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Who Yields in Orbit? Autonomous Satellites Need a Right-of-Way Rule,” his sixth original contribution for the outlet and his second commentary on space-domain governance. The piece opens with the European Space Agency’s September 2019 collision-avoidance decision, taken with almost no framework, and argues that low Earth orbit is approaching a regime of continuous conjunctions, with the largest constellation alone seeking to grow from twelve thousand satellites toward forty-two thousand. Its central claim is an authority-allocation gap: if the first collision between two autonomous, maneuver-capable satellites occurs under the present non-system, both operators will honestly insist they had priority, and neither will be wrong, because no rule ever established who did. Invoking the maritime precedent, in which the sea drew its rules of the road before traffic became unmanageable, the piece argues that orbit retains that chance and that the window is closing at eight kilometers a second. The argument extends the applicant’s research on authority allocation for autonomous systems into space traffic coordination, and follows one day after his three placements of July 21, 2026.

Read on Eurasia Review ↗

Published Commentary July 2026

Op-ed published in Eurasia Review

Eurasia Review · July 21, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant’s opinion piece “Why We Punish the Protection That Works,” his fifth original contribution for the outlet. The piece names the preparedness paradox: successful prevention erases the evidence of its own necessity, leaving publics to weigh visible costs against invisible benefits and to judge protection, after the fact, as overreaction. It traces the pattern from the Y2K remediation effort (estimated at 300 to 600 billion dollars and followed within weeks by a hoax verdict) through ozone-layer recovery and epidemic prevention, frames the underlying oscillation with Max Weber’s distinction between an ethic of conviction and an ethic of responsibility, and applies it to the year’s loudest foreign-policy argument, the US strike on Iran’s nuclear program, where countable costs are debated while the foreclosed catastrophe can enter no exhibit. The prescription is evidentiary rather than partisan: preserve contemporaneous records of the perceived threat, the alternatives considered, and the basis for acting, so that later verdicts weigh the decision actually made under the uncertainty that existed. The argument carries the research program’s central discipline, contemporaneous assurance records for high-stakes decisions, into the domain of public judgment, and is his third published commentary dated July 21, 2026.

Read on Eurasia Review ↗

Published Commentary July 2026

Op-ed published by Geopolitical Monitor

Geopolitical Monitor · July 21, 2026 · Authored by Burak Oktenli

Geopolitical Monitor, a Canadian international-intelligence publication (ISSN 1927-3045), published the applicant's opinion piece “America’s Allies Should Test the AI They Plan to Depend On,” his third original analysis for the outlet. The piece examines the June 2026 export-control suspension of Claude Fable 5, in which a US directive suspended access for any foreign national within hours, allied users included, and access was restored roughly nineteen days later only after a Commerce Department review that no allied government performed or observed (a vetted set of American critical-infrastructure organizations regained access on June 26, with controls lifted June 30). From this episode it draws a narrow, structural lesson: allied access to frontier AI is conditional and revocable, and when it is revoked, readmission runs on evidence generated in another capital. The prescription is a standing national test-and-evaluation capability for adopted AI, a funded bench of people, test protocols, and red teams, rather than sovereign frontier models, together with allied pooling through shared test benches, common red-team protocols, and reciprocal recognition of one another's evaluations modeled on shared airworthiness findings, situated against AUKUS Pillar 2 and NATO decision-support adoption. The piece extends the applicant's earlier RealClearMarkets analysis of the same suspension episode into an allied-policy prescription, and is one of four public placements dated July 21, 2026, alongside two Eurasia Review analyses and the Springer journal article.

Read on Geopolitical Monitor ↗

Published Commentary July 2026

Op-ed published in Eurasia Review

Eurasia Review · July 21, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant's analysis “The Next Offset Won’t Be a Smarter AI, It Will Be a Governable One.” The piece is the clearest public articulation of the thesis underpinning the applicant's research program: that US military-AI strategy is organized around the wrong question, chasing model capability when the binding constraint is governable authority, meaning who or what is permitted to act, through which interface, and with what oversight. It argues that a model's outputs change nothing until they cross an interface, credential, network, or human action, so capability that cannot be governed is exposure rather than advantage, and it calls for an “authority offset” (permission control, verification, segmentation, and exercised human fallback) in place of a fourth capability offset, on the ground that capable models proliferate fastest through commercial markets, open research, and theft while the institutional and technical capacity for disciplined, verifiable, revocable deployment does not. Drawing on the 2024 CrowdStrike outage, the 2015 Ukrainian grid recovery through preserved manual fallback, and nuclear command-and-control (the Sandia National Laboratories always/never principle, and the June 2026 NSPM-11 revision directing an updated DoD Directive 3000.09), it locates the real risk in integration architecture and validation discipline rather than model intelligence. The argument is the direct conceptual foundation of the applicant's runtime authority-governance work, and this is his fourth original contribution published by Eurasia Review in July 2026.

Read on Eurasia Review ↗

Published Op-Ed July 2026

Op-ed published in the Washington Examiner

Washington Examiner · July 20, 2026 · Authored by Burak Oktenli

The Washington Examiner, a national American news and opinion publication, published the applicant's op-ed “We Guarded China’s Oil for Free. Trump Sent Them the Bill.” The piece reads President Trump's July 2026 announcement that the United States will be reimbursed for guarding the Strait of Hormuz as an overdue burden-sharing doctrine rather than a transit-fee mechanism, observing that Asia receives nearly 90 percent of the strait's crude, that China is its largest single customer, and that U.S. energy independence is what makes the reimbursement demand credible. It argues Congress should move the war supplemental together with pledges from Gulf producers and wealthy Asian consumers, citing the 1991 Gulf War cost-recovery precedent (allies made good on 94 percent of pledged contributions) and the still-extant Defense Cooperation Account at 10 U.S.C. 2608. The placement extends the applicant's published commentary into national energy-security and burden-sharing policy at a major American outlet.

Read on Washington Examiner ↗ Washington Examiner author page ↗

Published Commentary July 2026

Op-ed published in Eurasia Review

Eurasia Review · July 19, 2026 · Authored by Burak Oktenli

Eurasia Review published the applicant's analysis “Two Years After CrowdStrike, We Are Still Afraid of the Wrong AI Apocalypse,” timed to the second anniversary of the July 2024 CrowdStrike outage. The piece argues that the prevailing AI-catastrophe scenarios (a universal off-switch, an AI escaping into the power grid, an autonomous nuclear launch) rest on a single confusion: they mistake what a system can compute for what it is permitted to touch. It contends that the scarce resource converting computational capability into physical consequence is authority, gated by credentials, integrations, standing permissions, and physical authentication, rather than intelligence, and that the genuine risk is the steady erosion of manual fallback competence as critical systems are automated. Drawing on the CrowdStrike, 2015 Ukraine grid, and Colonial Pipeline incidents (each bounded and recovered through human manual operation), it prescribes human-approval gates on consequential automated actions, an explicit inventory of what automation is permitted to touch as the real risk surface, and maintained, drilled manual procedures treated as critical infrastructure. The argument is a direct public articulation of the capability-versus-authority premise underlying the applicant's runtime authority-governance research, and is his third original contribution published by Eurasia Review in July 2026.

Read on Eurasia Review ↗

Published Commentary July 2026

Op-ed published by Geopolitical Monitor

Geopolitical Monitor (Situation Reports) · July 17, 2026 · Authored by Burak Oktenli · Syndicated by Eurasia Review on July 18, 2026

Geopolitical Monitor published the applicant's analysis “The Reciprocity Inversion: How Ukraine Became the Gulf’s Counter-Drone Lifeline” in its Situation Reports section. The piece traces Ukraine's 2026 deployment of counter-drone specialists across five Gulf partners, hardening by July into ten-year military-technical cooperation agreements with the UAE, Qatar, and Saudi Arabia, and argues the episode marks a reciprocity inversion in alliance architecture: the world's largest recipient of security assistance becoming, in one operationally decisive domain, the supplier of last resort, extending to Pentagon talks to purchase Ukrainian-made interceptor drones and a U.S. license for Ukraine to manufacture Patriot missiles. It grounds the inversion in cost-exchange arithmetic (roughly $35,000 attack drones against $4 million Patriot and $12.7 million THAAD interceptors, versus $800 to $3,000 Ukrainian interceptor drones producible at 2,000 a day) and in combat-derived calibration against a drone threat that iterates roughly every six weeks, which the piece identifies as a non-transferable strategic asset. The analysis extends the applicant's counter-UAS and attritable-systems research thread and is his second Geopolitical Monitor publication. Eurasia Review syndicated the analysis on July 18, 2026, its second syndication of a Geopolitical Monitor piece by the applicant.

Read on Geopolitical Monitor ↗ · Eurasia Review syndication ↗

Published Commentary July 2026

Analysis published in Eurasia Review

Eurasia Review · July 16, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant's analysis “What Six Years of Drone Proliferation Means for the Next Fight.” It traces armed-drone diffusion from the 2020 Nagorno-Karabakh war, where inexpensive strike drones and loitering munitions destroyed tanks, artillery, and air-defense systems, through Ethiopia, Sudan (over 1,000 drone attacks since April 2023), the Sahel, the Democratic Republic of the Congo, and Myanmar, now third globally for drone operations. The piece argues that the economics of attritability (systems far cheaper than the targets they destroy) and rapid mutation (fiber-optic control links defeating radio-frequency jamming, anti-jamming modules migrating into improvised fleets) make the primary threat to any deploying force a cheap, copied system rather than a peer arsenal. Citing U.S. Army doctrine (ATP 3-01.81) that pushes the counter-drone burden to the smallest tactical units, it concludes that low-cost adversary air must be treated as a baseline condition of every deployment, institutionalized through home-station training and standing small-unit battle drills. The analysis complements the applicant's BLADE-CUAS counter-UAS governance work and is his second original analysis published by Eurasia Review in July 2026.

Read on Eurasia Review ↗

Published Op-Ed July 2026

Op-ed published in RealClearDefense

RealClearDefense · July 16, 2026 · Authored by Burak Oktenli

RealClearDefense published the applicant's analysis “Military AI - The Confidence You Cannot See.” It argues that the documented pattern of operators deferring to machine confidence scores (and its mirror image, abandoning a system after one visible error) is conventionally treated as human psychology to be trained away, when it actually marks an unchosen migration of decision authority to machines while records certify that a human remained in charge. The piece explains that a confidence score is a byproduct of pattern matching rather than judgment, and that a bare percentage conceals whether the case at hand lies inside the range the system was validated against. It proposes treating calibration as part of the authority architecture: confidence communicated bound to its validated context, and the system's autonomous authority contracting toward the human when context drifts or calibrated confidence falls below what the stakes demand, an enforceable specification for procurement standards and audit rather than a demand on operator willpower. This is the applicant's third RealClearDefense publication. It was subsequently distributed through the RealClearWire syndication network and republished by outlets including The Ohio Press Network (the second of the applicant's op-eds that outlet has carried) and the KHQ NonStop Local broadcast group.

Read on RealClearDefense ↗ Ohio Press Network republication ↗ KHQ republication ↗

Published Commentary July 2026

Analysis published in Eurasia Review

Eurasia Review · July 14, 2026 · Analysis designation · Authored by Burak Oktenli

Eurasia Review published the applicant's analysis "The Governance Gap Hiding in Plain Sight Across Eight Industries." Prompted by the May 2026 Five Eyes joint warning on autonomous AI agents, it argues that eight sectors, from water utilities, aviation, and road transport to maritime, critical infrastructure, agentic finance, and defense, face one structural problem: autonomous systems now act faster than any human can review, permit, or reverse. It contends that regulating model capability and trusting developer guardrails both miss this temporal gap, and proposes a common governance layer at the point of action and outside the model, resting on three properties: permission evaluated by a component the system cannot overrule, human confirmation that scales with the stakes, and tamper-evident audit records, all buildable on existing NIST and EU frameworks. The piece extends the author's cross-domain research on runtime authority governance from defense to civilian and safety-critical sectors.

Read on Eurasia Review ↗

Published Op-Ed July 2026

Op-ed published in IT Ops Times

IT Ops Times (D2 Emerge) · July 13, 2026 · Authored by Burak Oktenli

IT Ops Times published the applicant's analysis “Your AI Agent’s Permission Model Is Built on a Boolean. Irreversible Actions Need More Than That.” It argues that production AI agents inherit a firewall-style allow/deny permission model whose validity rested on reversibility, and that the model breaks once agents take irreversible actions (moving money, disconnecting hosts, shipping configuration) faster than any human can review. Drawing on runtime-enforcement theory (Fred Schneider’s characterization of safety properties, and Jay Ligatti’s edit automata whose corrective power presupposes reversibility), it proposes three departures: graded authority as a continuous runtime value that rises with corroborating evidence and falls with anomalies, conservative failure so authority contracts under uncertainty, and an external interlock placing the final gate for irreversible actions outside the agent’s own compute path and trust domain. The argument maps the author’s runtime authority-governance research directly onto production AI-agent engineering.

Read on IT Ops Times ↗

Published Op-Ed July 2026

Op-ed published in RealClearMarkets

RealClearMarkets · July 13, 2026 · Authored by Burak Oktenli, listed on the RealClearMarkets author page

RealClearMarkets published the applicant's op-ed "The Cruel Economics of the 19-Day Anthropic Ban," an economic analysis of the June 2026 Commerce Department order restricting access to Anthropic's two most capable AI models. The piece argues that the restriction functioned as an unintended experiment in the economics of technology restriction: the denial was brief because near-substitute models exist, its costs fell on compliant banks, universities, and cyber-defenders rather than the intended target, and unpredictable access gives enterprises a durable incentive to diversify away from American AI providers. It recommends that future controls be designed at the level of specific actions against specific targets rather than an entire model's availability. The argument extends the author's research on the governance of authority in AI-enabled systems into export-control economics.

Read on RealClearMarkets ↗

Published Commentary July 2026

Op-ed published by Geopolitical Monitor, syndicated by Eurasia Review

Geopolitical Monitor · July 2026 · Syndicated by Eurasia Review on July 13, 2026

Geopolitical Monitor, an open-source intelligence and geopolitical risk analysis outlet, published the applicant's op-ed "Who Fires the Shot? Closing the Authority Gap in Indo-Pacific Autonomous Warfare," subsequently syndicated by Eurasia Review. The piece argues that Indo-Pacific coalitions, including AUKUS Pillar 2 and Quad-composition maritime operations, are fielding autonomous systems faster than members can agree on who may authorize them to act, and proposes a Coalition Authority Interoperability Protocol: computable national authority profiles, dependency caps, cryptographic multi-signature rules, and an auditable accountability layer, under which aggregate coalition authority never exceeds what its most conservative participating member permits. The argument applies the author's research on runtime authority governance to coalition interoperability. On July 16, 2026, Indian Strategic Studies, a strategic-affairs portal curated by Maj Gen P K Mallick, VSM (Retd), formerly Senior Directing Staff (Army) at India's National Defence College, featured the piece with a curated summary and an editorial comment engaging the proposal from an Indian strategic-autonomy perspective, the second of the applicant's op-eds that portal has carried.

Read on Geopolitical Monitor ↗ · Eurasia Review ↗ · Indian Strategic Studies feature ↗

Published Op-Ed July 2026

Op-ed published in RealClearDefense

RealClearDefense · July 2, 2026 · Authored by Burak Oktenli, listed on the RealClearDefense author page

RealClearDefense, a national-security commentary outlet, published the applicant's op-ed "The One Clause in Trump's AI Memo That Could Encumber the Off-Switch." It examines National Security Presidential Memorandum 11 (June 5, 2026) and warns that its "Reliability and Sovereign Control" mandate, intended to stop commercial AI vendors from disabling fielded military systems, could, if implemented without care, also lock out the human operator's ability to terminate a malfunctioning autonomous system, which is the core requirement of appropriate human judgment under DoD Directive 3000.09. It proposes that the mandated 90-day revision of Directive 3000.09 add an explicit operator-override exemption distinguishing authenticated human command interrupts, hardware-rooted termination, recovery fail-safes, and latency-bounded deliberation windows from the external vendor modifications the memorandum prohibits, and it calls for congressional oversight through the FY27 National Defense Authorization Act, including mapping the revised standards to the NIST AI Risk Management Framework and testing termination pathways in denied environments. The argument extends the author's research on runtime authority governance for autonomous and AI-enabled systems.

Read on RealClearDefense ↗

Published Commentary June 2026

Op-ed published by The Defense Post

The Defense Post · June 30, 2026 · Authored by Burak Oktenli, who also holds a curated contributor profile in the outlet’s member directory, listed as an author at The Defense Post

The Defense Post, a defense news and commentary outlet, published the applicant's commentary "When the Machine Decides and the Human Signs." It argues that human-on-the-loop oversight of AI-enabled targeting, as currently practiced, does not hold at high operational tempo: once a targeting pipeline produces recommendations faster than a commander can independently verify them, the effective decision shifts into the software while authority remains nominally with the human. Drawing on the 2026 Operation Epic Fury campaign as a case study, it argues for engineering oversight into the system rather than treating human approval as a safeguard the human cannot provide, and proposes automated checks on the currency of targeting intelligence, automatic pauses inside the software when anomalies appear, operational tempo tied to verification capacity, an updated U.S. Department of Defense Directive 3000.09 for mass-engagement scenarios, and congressional reporting on the review time commanders actually have. The argument extends the author's research on runtime authority governance for autonomous and AI-enabled systems.

Read on The Defense Post ↗

Published Commentary June 2026

Op-ed published by the Modern War Institute at West Point

Modern War Institute at West Point · June 26, 2026 · Authored by Burak Oktenli, listed as an author at the Modern War Institute

The Modern War Institute at West Point, a research center at the U.S. Military Academy, published the applicant's op-ed "When the Machine Acts First: Closing the Authority Gap on the Autonomous Battlefield." It argues that the collapsing distance between a machine's reasoning and its action opens an authority gap between the authority a commander holds and the authority a fielded autonomous system exercises before anyone can intervene, and proposes runtime authority governance: enforcing permission outside the model, pausing before consequence and failing toward inaction, and tying autonomy to the criticality of the target. The argument applies the author's research on runtime authority governance for autonomous and AI-enabled systems.

Read on Modern War Institute ↗

Editorial Recognition June to September 2026

Selected four times for UPI's Korea Regional Review Director's Corner

United Press International (UPI), Korea Regional Review · Director's Corner, National Security News & Commentary · June 26, August 19, September 3 and September 22, 2026

United Press International selected the applicant's work four times for the Korea Regional Review Director's Corner, an edited national-security review curated by the publication's Executive Director in which each selected item carries an accompanying editorial commentary. In June the review carried his Modern War Institute op-ed on closing the authority gap on the autonomous battlefield, alongside analyses from Foreign Affairs, The Wall Street Journal, the Financial Times, War on the Rocks, and Small Wars Journal, with the editor's commentary engaging the argument directly and framing the central challenge as one of command authority rather than technology, noting that autonomous systems must operate under enforceable human command so that speed does not outrun accountability and the law of armed conflict. In August the review selected his second Modern War Institute analysis, on resilience and the governance of autonomous systems, on the day it was published, placing it fifteenth among eighteen items alongside The Economist, The Telegraph, The Washington Post, Reuters, CNN, National Review, Foreign Affairs, War on the Rocks, and Small Wars Journal. The accompanying commentary did not summarise the piece but adopted and extended its recommendation, stating that resilience cannot mean automatic restoration of lethal authority, that adversaries will attack the decision architecture itself by manipulating sensors and communications to influence when autonomous systems regain permission to engage, and that autonomous weapons must therefore be designed, tested and red-teamed for adversarial manipulation of their decision processes rather than only for their ability to survive jamming. A curator returning to an author's work eight weeks later, and converting his proposal into a stated requirement, is a stronger indicator than a single selection, and the same-day timing shows the analysis was read on publication. On September 3, 2026 the same editor selected a third piece, the applicant’s RealClearDefense analysis of cumulative public disclosure, placing it as item twenty-two in a twenty-three item national-security digest whose other entries were drawn from Foreign Affairs, the Wall Street Journal, the New York Times, the Washington Post, War on the Rocks, the Diplomat and Small Wars Journal. The accompanying editorial comment restates the argument in the editor’s own terms, that the issue is not secrecy versus transparency but the intelligence value of aggregation, that adversaries can assemble public disclosures into estimates of capacity, timing, vulnerabilities and dependencies, and that an adversary-view mosaic test should be applied to high-value releases without creating another reflexive classification bureaucracy. On September 22, 2026 the same editor selected a fourth piece, the applicant’s RealClearDefense analysis of target-data assurance, placing it as item twenty-one in a twenty-six item national-security digest whose other entries were drawn from The New York Times, Foreign Affairs, The Wall Street Journal, The Atlantic, Newsweek, War on the Rocks, Small Wars Journal and a Washington defence research centre. The accompanying comment again restates the argument rather than summarising it, holding that speed without accurate information is not precision but faster error, that machine assistance can carry outdated intelligence into a present-day strike, that adding another person to the loop is no remedy unless that person has information which is current, traceable and open to challenge together with the authority to stop the process when reality no longer matches the database, and that the requirement is to compress the sensor-to-shooter timeline without compressing the time needed to establish what is actually being struck. Selection into that company, four times by the same editor, is a judgement about which analysis a professional national-security readership should read that day.

Review, August 19, 2026 ↗ Review, June 26, 2026 ↗ Third selection, September 3, 2026 ↗ Fourth selection, September 22, 2026 ↗

Editorial Recognition June 2026

Featured in the Parliamentary Observatory on AI of the Parliamentary Assembly of the Mediterranean

Centre for Global Studies, Parliamentary Assembly of the Mediterranean · Weekly Digest on AI and Emerging Technologies · June 23, 2026

The Centre for Global Studies of the Parliamentary Assembly of the Mediterranean, an international inter-parliamentary institution, selected the applicant's RUSI commentary "The Hour That Worked: What Midnight Hammer Teaches About AI-Era Command" for its Parliamentary Observatory on AI weekly digest. The commentary was summarized in the digest's Defense, Intelligence, and Warfare section alongside analyses from Defense One, Defense News, and Breaking Defense, carrying the applicant's argument that command authority designed into a mission in advance offers a model of disciplined command for an era in which machines compute faster than humans can deliberate.

Read on the Parliamentary Assembly of the Mediterranean ↗ · Centre for Global Studies mirror ↗

Published Commentary June 2026

Op-ed published by The Space Review (in association with SpaceNews)

The Space Review, in association with SpaceNews · June 22, 2026 · Authored by Burak Oktenli · First of his three placements with the outlet · Names the authority architecture

The Space Review, a space-policy publication issued in association with SpaceNews, published the applicant’s op-ed “Space Autonomy Needs an Authority Architecture Before 2027,” the first of his three placements with the outlet. He opens on a hypothetical early-warning satellite in high orbit that detects a co-orbital object closing fast, where the light-time delay to the ground and the time needed to analyse, seek approval and transmit a command exceed the closing window, and asks the question the piece exists to answer: WHEN REAL-TIME HUMAN CONTROL BECOMES PHYSICALLY IMPOSSIBLE, WHO MAKES THE CALL? His framing of why space differs is exact. On land, at sea and in the air, keeping a person in the decision loop remains both the preferred doctrine and a feasible standard; in orbit, SIGNAL LATENCY TRANSFORMS AUTONOMY FROM AN OPERATIONAL CONVENIENCE INTO A PHYSICAL NECESSITY, and when adversarial jamming severs the link, the delay stretches into practical infinity, so that in those dark windows autonomy is the only option the spacecraft has. *** This is the piece that names the AUTHORITY ARCHITECTURE. *** Having observed that the military space service’s foundational doctrine defines orbital and electromagnetic warfare but does not address the authority of autonomous systems, and asked who holds legal and operational authority when an autonomous vehicle decides to approach or intercept, and who bears responsibility if that action sparks a crisis, he defines what is missing: an institutional spine, embedded in both military culture and system software, that states WHAT DECISIONS A MACHINE MAY MAKE UNDER WHICH CONDITIONS, THE LEGAL BASIS FOR THOSE DECISIONS, AND HOW ACCOUNTABILITY TRACKS BACK UP THE CHAIN OF COMMAND. An independent Indo-Pacific strategy service later carried this piece under a headline of its own using exactly that term, recorded separately on this page. His timing argument is a design-time claim: demonstrations are moving toward operational missions and force design is on the drafting table now, so if governance, delegation and accountability are not embedded in hardware and software during the design phase, RETROFITTING THEM INTO BILLION-DOLLAR SYSTEMS LATER WILL FAIL TECHNICALLY. His three-step roadmap follows. Doctrine should define the thresholds beyond which a system may decide independently AND the attribution of responsibility for erroneous decisions, in accordance with the chain of command. Governance should be treated as AN ENGINEERING REQUIREMENT, NOT MERELY A LEGAL MATTER, with vehicles built from the outset to constrain and audit their own decisions. And where delay or jamming makes human approval impossible, every autonomous decision should be recorded and, once communication is restored, made auditable through a transparent, immutable, cryptographically secured DEFERRED ACCOUNTABILITY record. *** The deferred accountability record is the earliest form of an instrument he developed across the following months. *** Three months later, in the same outlet, the idea became the MANEUVER RECEIPT with its nine specified fields, and in a later national outlet it became the decision record he proposed for high-consequence manoeuvres by armed spacecraft. The progression from a principle in June to a field-level specification in September is visible within his own published record. *** The sentence that carries his whole programme appears here first in this outlet. *** WE ARE FORCED TO DELEGATE OPERATIONAL INDEPENDENCE TO WEAPONS AND SATELLITES, BUT WE CANNOT DELEGATE STRATEGIC AUTHORITY AND COMMAND RESPONSIBILITY. That separation, between what a machine may do and who remains answerable for it, is the line he later compressed as a tactical optimisation that must not quietly become a strategic signal, and it runs through every domain he has since written about. His statement of the question he wants asked is the other line worth carrying: the issue is not whether autonomous systems will exist in space, but WHETHER THEY WILL BE ACCOUNTABLE, TRANSPARENT AND CONTROLLABLE. The closing refers to his own research framework by name, which is recorded here as a self-reference. The outlet’s biography on this piece states his graduate study in its correct in-progress form. No official, service, command, company, vehicle or doctrine document is identified by name in this record.

Read on The Space Review ↗

Published Commentary June 2026

Guest commentary published by RUSI (Royal United Services Institute)

Royal United Services Institute (RUSI) · June 16, 2026 · Authored by Burak Oktenli, listed as a Guest Contributor at RUSI

RUSI, one of the world's oldest defense and security institutes, published the applicant's guest commentary "The Hour That Worked: What Midnight Hammer Teaches About AI-Era Command." The piece argues that the authority to apply force, and the validation of the data behind it, should be established at an inspectable point before deployment rather than improvised at machine speed, contrasting Operation Midnight Hammer with the AI-accelerated Operation Epic Fury and drawing implications for NATO interoperability and national "red card" authority. This is authored analysis placed at a leading international defense institution, directly aligned with the applicant's research on authority governance for autonomous and AI-enabled systems. Read on RUSI. The commentary was also republished at MERO (July 18, 2026). An Italian security and defense news service, GrNet, published an Italian summary of the analysis on June 16, 2026. Shared by RUSI's official social channels (June 16, 2026): view post.

Published Op-Ed June 2026

Published commentary in RealClearDefense

RealClearDefense · June 4, 2026 · Authored by Burak Oktenli

RealClearDefense, a national-security commentary outlet, selected and published the applicant's op-ed "The Quantum Clock Is Already Ticking on America's Autonomous Arsenal," which argues that post-quantum cryptography is a command-authority problem for autonomous weapons, not only a secrecy problem. This is authored commentary placed in a recognized outlet, distinct from the third-party citations above. It was subsequently featured and summarized in the Defense Tech and Acquisition newsletter (MacGregor and Modigliani), June 2026, and picked up by the Indian Strategic Studies digest (Maj Gen P K Mallick, VSM Retd) the same month. Read on RealClearDefense. RealClearDefense author page. Read on RealClearWire. RealClearWire author page. The Ohio Press Network. Eastern Progress.

Published Commentary June 2026

Published commentary in Military AI

Military AI (militaryai.ai) · June 12, 2026 · Authored by Burak Oktenli

Military AI, a defense-AI commentary publication, published the applicant's article "When an AI Agent Fires First, Who Answers for It?", which argues that traceable human authority must be a design requirement for agentic military systems before they are fielded. The piece engages the Five Eyes "Careful Adoption of Agentic AI Services" guidance (CISA, NSA, and partners, 2026). This is authored commentary placed through an editor at a recognized defense outlet. Read on Military AI.

Academic Citation May 2026

Cited in "The Cascade Protocol"

Joey Hernandez · Pomona College, Department of Economics · ISBN 979-8195946579

A published academic book on cascading AI failures in military and strategic decision-making cited the AI-Enabled Military Decision-Making preprint as Reference #36 of 36. The book sits the work alongside the U.S. Department of War AI Strategy (January 2026), King's College London, US Army War College, Harvard Law School PILAC, Lieber Institute West Point, Carnegie Endowment for International Peace, Atlantic Council, and NDU Press. Independent third-party adoption of the applicant's research within four months of publication, with full ISBN registration and DOI minting on Zenodo.

Work Cited: Oktenli, B. (2026). "AI-Enabled Military Decision-Making and Escalation Risk: Human-Machine Command Authority in Great Power Competition." SSRN ID 6082847.

View Citing Work on Zenodo View Cited Paper on SSRN

Academic Citation April 1, 2026

Cited in "The Ethics of AI in U.S. Warfare"

LexAI Journal, University of Toronto Law and Ethics of Artificial Intelligence Students' Association

A multi-author academic article published by the University of Toronto's Law and Ethics of Artificial Intelligence Students' Association cited the ERAM framework paper as Reference #4. The article examines decision compression in AI-enabled military command, directly referencing the ERAM framework's analysis of how automated decision-support systems destabilize human oversight in military operations.

Work Cited: Oktenli, B. (2026). "Decision Compression and Escalation Risk in AI-Enabled Military Command and Control: An Operational Analysis of the ERAM Framework." SSRN.

View Source Article

Academic Citation 2025

Cited in Defense Logistics Paper Addressed to Ministries of Defense

Jorge Enrique Rivera Rojas, Biologist & Freeze-Drying Consultant · Bogotá, Colombia

A bilingual (English/Spanish) strategic paper addressed to Ministries of Defense, Armed Forces, and state institutions across multiple countries cited the LinkedIn article as a strategic analysis source, listed alongside official U.S. government sources including the Department of Defense (DoD), Defense Logistics Agency (DLA), U.S. Army DEVCOM Soldier Center, and USDA.

Work Cited: Oktenli, B. "Why Freeze-Dried Food Is Becoming a Strategic Asset in Defense, Emergency Preparedness, and National Security." LinkedIn Pulse.

View Full Paper (PDF) Academia.edu

Academic Citation June 2026

Cited in "Cognition, Agency, and Authority: A Taxonomy of Advanced AI Systems"

Lt Gen (Dr) R S Panwar (Indian Army, Retired) · Doctorate in Computer Science, Indian Institute of Technology Bombay · Academia.edu preprint

A defense-technology preprint by Lt Gen (Dr) R S Panwar, a retired Indian Army officer and computer-science doctorate from IIT Bombay, cites the applicant's working paper on AI-enabled military decision-making and escalation risk (endnote 39) in support of its argument that cognitively sophisticated but non-agentic AI systems carry under-recognized strategic risk. The citing work develops a six-class taxonomy of advanced AI systems organized around authority delegation and command responsibility, themes directly adjacent to the applicant's research on human-machine command authority.

Work Cited: Oktenli, B. (2026). "AI-Enabled Military Decision-Making and Escalation Risk: Human-Machine Command Authority in Great Power Competition." SSRN ID 6082847.

View the citing work ↗ · Download PDF ↗

Platform Feature 2026

Featured by SSRN Community

SSRN Official Community, Facebook (Elsevier / Social Science Research Network)

The official SSRN Community page, operated by Elsevier's Social Science Research Network, one of the world's largest open-access academic repositories, selected and featured the submarine surface signatures paper on their social media platform, amplifying its visibility to SSRN's global research audience.

Work Featured: Oktenli, B. "Physics-Based Analysis of Submarine Surface Signatures: Hydrodynamic Mechanisms and Detection Frameworks." SSRN.

SSRN Community

Official SSRN / Elsevier · Facebook

Physics-Based Analysis of Submarine Surface Signatures: Hydrodynamic Mechanisms and Detection Frameworks

Featured by the official SSRN Community page, operated by Elsevier's Social Science Research Network, which amplified this research to their global academic audience.

View Original Post on Facebook

View Post

Professional Community January 4, 2026

Shared by Wargaming Weekly

Wargaming Weekly, Bluesky (@wargamingweekly.bsky.social)

Wargaming Weekly, a defense and military simulation community account, independently shared the LinkedIn article on AI-driven simulation in U.S. defense testing with the quote: "By injecting AI agents, massive parallelism, and data-driven analysis into simulations, the military can explore far more scenarios, catch hidden flaws, and train personnel more effectively."

Work Shared: Oktenli, B. "AI-Driven Simulation: Transforming US Defense Testing." LinkedIn Pulse.

View Post

Imprint 2026

Authority & Architecture: Publishing Imprint of Record

Authority & Architecture Imprint · Established 2026

Ten-volume technical reference series across three book trilogies, distributed via Amazon and IngramSpark. All volumes assigned ISBNs registered through Bowker. Library of Congress Control Numbers assigned to each series: The Authority Equation 2026912260, Autonomous Authority 2026922204, and The Authority Discipline 2026922201.

View Imprint Catalog

Media Feature

Turkish Media Appearance

Haberler.com, Dailymotion

Featured in Turkish-language media coverage on Haberler.com, one of Turkey's major digital news platforms.

View Video

Peer-Reviewed Publication August 2026

Second peer-reviewed journal article published, in a second discipline

International Journal of Contemporary Security Studies · Faculty of Security Studies, University of Belgrade · Vol. 2, No. 2 (2026) · Published online First August 24, 2026

The International Journal of Contemporary Security Studies, a double-blind peer-reviewed open-access journal published by the Faculty of Security Studies at the University of Belgrade, published the applicant’s article “The Ingratitude Paradox: A Causal Mechanism of Democratic Discounting of Successful Protection.” The article appears at pages 55 to 68 of the issue, carries the journal’s review-article designation, records submission on 18 June 2026, revision on 15 July and acceptance on 25 July, and is published under the applicant’s Georgetown University affiliation; its digital object identifier is to be registered when the issue is finalised in December 2026. This is his second article in a peer-reviewed journal and the first outside engineering, following the sensor-attestation protocol published in Springer’s Journal of Hardware and Systems Security in July. The article is a theory-building and mechanism-specification study rather than an empirical test, and it says so explicitly. It specifies a five-stage causal mechanism: a threat generates demand for protection; the protective action imposes visible procedural costs; the action succeeds and eliminates the threat; the eliminated threat becomes perceptually invisible; and the visible cost is then weighed against the now-invisible benefit, producing a verdict that the protection was unnecessary or disproportionate. Its defining property is endogeneity, since the threat-eliminating step is also the invisibility-generating step, which makes the tendency intrinsic to protective success rather than a contingent failure of communication. The argument is grounded in the securitization literature and the availability-heuristic tradition in risk perception, is differentiated systematically from adjacent concepts, is shown to be symmetric with respect to warrant and therefore explanatory rather than justificatory, is illustrated through a documented positive case traced across all five stages, and is bounded by scope conditions established through a contrastive case in which the mechanism did not fire. The article states a falsifiable prediction relating the depth of the discounting verdict to the completeness of threat-elimination, and identifies that prediction as untested with its systematic evaluation named as the natural next step. The publication demonstrates that the applicant’s central concern, how institutions evaluate protective systems whose benefits are not directly observable, generalises beyond the engineering domain into security theory, and it converts an argument he had previously made in the commentary press into a specified and falsifiable scholarly contribution.

Read the article ↗

Professional Service October 2026

Invited to peer review for the International Journal of System Assurance Engineering and Management, and accepted

International Journal of System Assurance Engineering and Management (Springer Nature; ISSN 0975-6809) · Invited by the journal’s lead guest editor and accepted, October 13, 2026 · Review due November 18, 2026

The lead guest editor of the International Journal of System Assurance Engineering and Management, a Springer Nature journal co-published with the Society for Reliability Engineering, Quality and Operations Management in India and the Division of Operation and Maintenance at Luleå University of Technology in Sweden, invited the applicant on October 13, 2026, in view of his expertise, to review a manuscript submitted to the journal, and he accepted the assignment the same morning. The journal confirmed the assignment that day and set November 18, 2026 as the date for his review. System assurance engineering, the journal’s subject, covers the reliability, safety, quality and risk of large engineered systems, the field to which his own research on the assurance of autonomous and AI-enabled systems belongs. It is his first reviewer assignment for a Springer Nature journal and brings his peer-review service to five journals, with the Journal of Responsible Technology, F1000Research, Connections: The Quarterly Journal and IEEE Engineering Management Review. The review is not yet filed, and this entry will be updated when it is completed. The manuscript’s title, authors and content remain confidential to the review process and are not disclosed in this record.

Journal homepage ↗

Professional Service October 2026

Second published peer review for F1000Research, approving the revised article

F1000Research · Review of version 4 published October 10, 2026 · Report DOI 10.5256/f1000research.211354.r528512 · Follow-up to his August report on the same article

F1000Research published a second report by the applicant, under his own name and with its own digital object identifier, on the article he first reviewed in August, a comparative study of algorithmic targeting systems and their compliance with international humanitarian law. The author revised the article into a fourth version, published October 9, 2026, with a point-by-point response to his first report, and the applicant reviewed the revision against that response. He found that all six major concerns from his first report had been addressed in a substantive and satisfactory manner: the numerical scoring of legal compliance replaced by a qualitative classification with explicit inferential limits; the unsupported extrapolation from a reported error rate to a number of misidentified individuals removed; an evidentiary-confidence level attached to each assessment, with absence of evidence distinguished from low capability; the mismatch between the analytical framework and its cited source corrected; the legal review of new weapons separated from the concept of meaningful human control; and the bibliographic and source-traceability problems corrected. He also credited the revised taxonomy, the comparison logic within each category, and the separation of technical accuracy, operational effectiveness, legal compliance and humanitarian outcome with materially improving the coherence of the paper. He left two minor residual comments, one asking that a sentence treating one side’s capability as proven absent be restated so that the absence of public evidence is recorded as undetermined, consistent with the paper’s own revised rule, and one asking that the final sentence of the conclusion be rephrased around the paper’s volume and time argument, and he changed his decision from approved with reservations to approved. The report carries a competing-interests statement recording that he has published related commentary, has no personal, financial, institutional or collaborative relationship with the author, and had disclosed the subject-matter overlap to the editorial team. The article record now shows one approval and four approvals with reservations among its five reviewers. This report belongs to the same review assignment as his August report and is not counted as a separate journal engagement.

Read the published review ↗ Report DOI ↗ Article record ↗

Professional Service September 2026

Peer review completed for IEEE Engineering Management Review, certified by the Editor-in-Chief

IEEE Engineering Management Review, a journal of the Institute of Electrical and Electronics Engineers · Invited by a department editor and accepted, September 22, 2026 · Completed ahead of the October 6 due date · Certificate of Reviewer Recognition issued September 30, 2026

A department editor of IEEE Engineering Management Review invited the applicant to review a manuscript submitted to the journal, and he accepted the assignment the same morning. He completed the review ahead of its October 6, 2026 due date, and on September 30, 2026 the journal issued him an IEEE Certificate of Reviewer Recognition, signed by its Editor-in-Chief, certifying that he has reviewed the manuscript for publication in the journal. This is his third completed peer review, after his published open review for F1000Research and his review for Connections: The Quarterly Journal, and the first for a journal of the Institute of Electrical and Electronics Engineers. The manuscript falls within his field. The certificate identifies it only by the journal’s reference number; its title, authors and content remain confidential to the review process and are not disclosed in this record.

View the certificate ↗

Professional Service August 2026

Accepted as a technical program volunteer for IEEE ICC 2027

IEEE International Conference on Communications 2027 · Washington, DC · Technical Program and Workshops

Following an offer to contribute, the organizing leadership of IEEE ICC 2027 accepted the applicant as a volunteer in the technical program and workshops area and scheduled him into the conference’s in-person organizing meeting in September 2026. IEEE ICC is one of the two flagship conferences of the IEEE Communications Society, drawing close to two thousand attendees from more than seventy countries, and the 2027 edition is scheduled for Washington, DC in late May 2027. The technical program and workshops function covers paper review, workshop organization, and related program work, so the role applies the applicant’s referee-level familiarity with formal verification, systems security, and autonomous systems to the selection of work presented at a flagship IEEE venue. The specific committee assignment is settled at the organizing meeting; recorded here at the level the acceptance establishes.

IEEE ICC 2027 ↗

Professional Service August 2026

Peer review completed for Connections: The Quarterly Journal, confirmed in writing

Partnership for Peace Consortium of Defense Academies and Security Studies Institutes · Review conducted August 2026 for issue 25-1 · Completion confirmed in writing by the journal’s senior editor

The managing editor of Connections: The Quarterly Journal, the peer-reviewed journal of the Partnership for Peace Consortium of Defense Academies and Security Studies Institutes, invited the applicant to review a submitted article described as falling within his area of subject-matter expertise, and he accepted with a review due at the end of August. The review was completed and the journal confirmed it in writing on 24 August 2026. The confirmation was issued by the Chief of Academic Publications at the George C. Marshall European Center for Security Studies, who serves as the journal’s senior editor, and states that the applicant conducted an exhaustive and highly professional review for an article under consideration for issue 25-1. In subsequent correspondence the same editor described his review as by far the most comprehensive she had seen, which is an assessment of the quality of the work rather than of its completion, offered unprompted by an editor who handles a large volume of reviews. This is the third journal to engage him as a reviewer within a single month, following an acknowledged reviewer offer at an Elsevier journal and an assigned open review for a Taylor & Francis platform. Reviewing for a defense-education consortium journal places his assessment inside the professional literature of the field his research addresses, and the invitation was unsolicited. In accepting, the applicant recorded that he has no connection to the author, no prior knowledge of the manuscript, and no competing interest, and undertook to disclose immediately should the author’s identity become known to him. Details of the manuscript are withheld here consistent with reviewer confidentiality, and the journal operates a closed review model, so unlike his F1000Research report the assessment itself is not public; the written confirmation is the record of it.

Professional Service August 2026

Published peer reviewer for F1000Research (Taylor & Francis)

F1000Research · Review published August 11, 2026 · Report DOI 10.5256/f1000research.202374.r507419 · Named in the public reviewer roster on the platform’s AI and Machine Learning gateway

The applicant served as an invited peer reviewer for F1000Research, the open research publishing platform of the Taylor & Francis Group, and his review is now published in full under his own name alongside the article it assesses. F1000Research operates a fully open peer-review model in which the reviewer’s identity, complete report, and approval decision are published and independently citable; the report carries its own digital object identifier. The article reviewed is a comparative study of algorithmic targeting systems and their compliance with international humanitarian law, and the applicant returned a decision of approved with reservations. He is listed by name among the five peer reviewers displayed on the article record and on the platform’s artificial-intelligence and machine-learning gateway, where the article stood at its third version, with all five reviewers returning approved with reservations, when his report was published. That roster is a public, permanent and independently checkable record of the assignment, separate from the report itself. His report sets out six major comments in a structured form, each stating the issue, why it matters, the revision required, and the evidence needed to close it. Its substantive findings include that assigning numerical scores on a linear scale to inherently contextual legal principles conveys an appearance of methodological precision the underlying evidence cannot support and should be replaced with a qualitative classification; that a casualty extrapolation derived by applying an unverified error rate to an unverified target count gives speculative arithmetic greater evidentiary weight than its sources permit; that a comparison drawing on markedly asymmetric evidence should pair each assessment with an explicit evidentiary-confidence designation rather than presenting all scores as equally supported; and three sourcing corrections, namely an analytical framework attributed to a publication that is a book review rather than the origin of that framework, a definition of meaningful human control mis-sourced to a legal provision that in fact governs the review of new weapons, and a citation treating a publication platform as though it were an individual author. The report closes with an explicit statement of what the applicant was unable to assess, namely classified or non-public information on system architectures, training data, and operational performance, and records that his assessment is therefore confined to the quality and interpretation of the publicly available evidence cited. The published expertise attached to the report is military AI applications, autonomous weapon systems, international humanitarian law, operational targeting architecture, and defense technologies. This is independently verifiable evidence of the applicant’s standing as a reviewer of scholarly work in his field, and it is the second of three journal reviewer engagements accepted within a single month.

Read the published review ↗ Report DOI ↗ Listed on ORCID ↗ Article record ↗ Gateway listing ↗

Professional Service June 2026

Offered to peer review for the Journal of Responsible Technology

Journal of Responsible Technology (Elsevier, on behalf of ORBIT); ISSN 2666-6596

Offered to serve as a peer reviewer for the Journal of Responsible Technology; the Editor-in-Chief acknowledged the offer and circulated it to the journal's Associate Editors (June 2026).

Journal Homepage