# BLADE-FINANCE Governance Node (v1.0)

Authority governance for financial-sector AI decision systems under the U.S. Treasury
Financial Services AI Risk Management Framework (FS AI RMF).

Author: Burak Oktenli, Independent Researcher (AUTHREX Systems); Georgetown University,
M.P.S. Applied Intelligence. ORCID 0009-0001-8573-1667. Washington, DC.
License: Creative Commons Attribution 4.0 International (CC BY 4.0).
DOI: 10.5281/zenodo.20374692.
Website: burakoktenli.com. Project page: burakoktenli.com/blade-finance. Simulation: burakoktenli.com/blade-finance-simulation.

## What this deposit is

A simulation-validated, software-enforced authority-arbitration reference architecture for
financial-sector AI decision systems. An eight-stage AUTHREX pipeline (VALIDATE, SATA, ADARA,
MAIVA, HMAA, FLAME, ERAM, CARA) appends every decision to a SHA-256 evidence chain. A four-tier
HMAA authority model routes a transaction to autonomous clearance, supervised review, elevated
confirmation, or manual hold. A population-state model scores coordination across account,
device, payee, and IP-cluster history, and a retrospective stigmergic swarm-review module
recovers coordinated low-and-slow rings that the per-transaction path clears.

## Maturity and scope (read before citing)

- TRL 3-4 (simulation) / TRL 2 (hardware). No prototype has been built.
- All data are synthetic. The work has not been deployed in any financial institution, and the
  author has not been engaged by any institution to provide AI-governance services.
- The reported recall is an actionable-risk triage measure, not an empirical fraud-detection rate.
- The retrospective swarm review claims ensemble agreement and ring detection, not Byzantine
  fault tolerance; no quorum-intersection safety bound is claimed.
- The simulator uses real SHA-256 over canonical-form serialization; per-record ECDSA P-256 and
  HSM key custody are design-specified, not exercised in the browser.

## Contents

| File | Description |
|------|-------------|
| blade-finance-governance-node.pdf | Research paper (17 pp): architecture, equations, simulation results, limitations, references. |
| blade-finance-simulation.html | Reference simulator v2.2. Deterministic (mulberry32), real SHA-256 over canonical-form serialization, schema validation, triage metrics with Wilson intervals, Monte Carlo runner, external-dataset benchmark, golden-trace export, nine self-tests, retrospective swarm-review module. Runs offline, client-side. |
| blade-finance_authority_node_PARTS.csv | Reference authority-node bill of materials (36 components, approximately $9,228). |
| blade-finance_authority_node_ELECTRICAL_CONNECTIONS.json | 33 electrical connections. |
| blade-finance_authority_node_MECHANICAL_CONNECTIONS.json | 32 mechanical connections. |
| blade-finance_authority_node_CONFIG.json | Node definitions and provenance metadata. |
| blade-finance_authority_node_SCHEMATIC.svg | Vector wiring schematic. |
| blade-finance_authority_node_render.svg | Reference-node block diagram. |
| zenodo.json | Deposit metadata for upload. |

## Reproducibility

The simulator is deterministic and runs offline in any modern browser with no network access.
Open blade-finance-simulation.html, run the in-browser Self-Test (expect 9 of 9 pass), then run
the Monte Carlo control on the mixed-stream scenario at seed 42 with governance enabled. Stepping
the deterministic stream to 1,000 records at seed 42 fixes the audit-ledger head hash to
7d5aaab4... and the running checksum to 0x06acd6be; any divergence indicates a modified engine or
a non-conforming SHA-256 implementation. Triage confidence intervals are Wilson score intervals
over the 2,000-trial run.

## How to cite

Oktenli, B. (2026). BLADE-FINANCE Governance Node (v1.0). Zenodo.
DOI 10.5281/zenodo.20374692.