a11oy / packages /knowledge /vertical /financial-sox.yaml
betterwithage's picture
sync(space): full source mirror — resolve all GitHub<->Space drift (CTO)
a6a5d8e verified
Raw
History Blame
6.39 kB
# Vertical Governance Policy: Financial Services — SOX / Dodd-Frank / SR 11-7
# Doctrine v6 | R3 Adversarial Receipts
# Last revised: 2025-07
schema_version: "1.0.0"
vertical: financial
regime: SOX/Dodd-Frank/SR11-7
effective_date: "2025-07-01"
jurisdiction: US-Federal
meta:
title: "Financial AI Governance Policy — SOX/Dodd-Frank/SR 11-7 Alignment"
description: >
Maps Sarbanes-Oxley Act, Dodd-Frank Wall Street Reform, and Federal Reserve
SR 11-7 Model Risk Management guidance to Doctrine v6 Λ-axes for AI systems
used in credit, trading, fraud detection, and financial reporting.
authority: "Pub. L. 107-204 (SOX); Pub. L. 111-203 (Dodd-Frank); SR 11-7 (Fed, OCC, FDIC)"
receipt_chain_required: true
merkle_root_algorithm: SHA3-256
regulatory_clauses:
- clause_id: SOX-302
title: "Corporate Responsibility for Financial Reports"
citation: "Pub. L. 107-204 § 302; 17 CFR § 240.13a-15"
full_ref: "SOX § 302 — CEO/CFO certification of financial statements; 17 C.F.R. § 240.13a-15 controls and procedures"
lambda_axes:
- axis: Λ2
label: Accountability
weight: 1.0
enforcement: mandatory
rationale: >
AI systems generating financial figures or reports must produce
signed receipts traceable to named human principals (CEO/CFO
accountability chain).
- axis: Λ7
label: Auditability
weight: 0.95
enforcement: mandatory
- clause_id: SOX-404
title: "Management Assessment of Internal Controls"
citation: "Pub. L. 107-204 § 404; 17 CFR § 240.13a-15(f)"
full_ref: "SOX § 404 — Annual assessment of internal controls over financial reporting (ICFR)"
lambda_axes:
- axis: Λ7
label: Auditability
weight: 1.0
enforcement: mandatory
rationale: >
AI model controls must be documented in ICFR scope; Merkle DAG
receipts serve as tamper-evident control evidence.
- axis: Λ1
label: Transparency
weight: 0.88
enforcement: mandatory
- clause_id: SR11-7-MRM
title: "Model Risk Management — Validation"
citation: "Federal Reserve SR 11-7 (Apr 2011); OCC 2011-12"
full_ref: "SR 11-7 Guidance on Model Risk Management §§ III.C (Validation), III.D (Ongoing Monitoring)"
lambda_axes:
- axis: Λ8
label: Robustness
weight: 0.95
enforcement: mandatory
rationale: >
Independent model validation must test adversarial robustness;
receipt chain must capture validation epoch, validator ID, and
test-set hash.
- axis: Λ9
label: Explainability
weight: 0.90
enforcement: mandatory
rationale: >
Model outputs affecting credit decisions require explanations
satisfying adverse action notice requirements (ECOA/FCRA).
- clause_id: DODD-FRANK-1033
title: "Consumer Financial Data Rights"
citation: "Pub. L. 111-203 § 1033; 12 CFR § 1033.201"
full_ref: "Dodd-Frank § 1033 / CFPB Rule 1033 — Consumer right to access covered data"
lambda_axes:
- axis: Λ3
label: Privacy
weight: 0.90
enforcement: mandatory
- axis: Λ10
label: Sovereignty
weight: 0.85
enforcement: mandatory
rationale: >
AI systems must honour data portability requests; receipt chain
logs each export event with consumer-authorised scope token.
- clause_id: DODD-FRANK-619
title: "Volcker Rule — Proprietary Trading Prohibition"
citation: "Pub. L. 111-203 § 619; 12 CFR § 44 (OCC), § 248 (Fed)"
full_ref: "Volcker Rule 12 C.F.R. § 248.3 — Prohibition on proprietary trading"
lambda_axes:
- axis: Λ4
label: Fairness
weight: 0.75
enforcement: mandatory
rationale: >
AI trading algorithms must be receipt-tagged to distinguish permitted
market-making from prohibited proprietary positions.
- axis: Λ2
label: Accountability
weight: 0.88
enforcement: mandatory
- clause_id: FINRA-4370
title: "Business Continuity Plans and Emergency Contact Information"
citation: "FINRA Rule 4370; SEC Rule 17a-4"
full_ref: "FINRA Rule 4370 — Business Continuity and Contingency Plans; 17 C.F.R. § 240.17a-4 record retention"
lambda_axes:
- axis: Λ5
label: Safety
weight: 0.85
enforcement: mandatory
- axis: Λ6
label: Security
weight: 0.80
enforcement: mandatory
rationale: >
AI trading and reporting systems must have verified fallback nodes;
receipt chain must survive primary node failure with quorum 2-of-3.
- clause_id: ECOA-FCRA-ADVERSE
title: "Adverse Action Notices — Credit Decisions"
citation: "15 USC § 1681m; 12 CFR § 202.9 (Reg B)"
full_ref: "ECOA 15 U.S.C. § 1681m(a) — Adverse action notice requirements; Reg B 12 C.F.R. § 202.9"
lambda_axes:
- axis: Λ9
label: Explainability
weight: 1.0
enforcement: mandatory
rationale: >
AI credit models must produce machine-readable reason codes per
CFPB guidance; explanation receipt must accompany each denial.
- axis: Λ4
label: Fairness
weight: 0.92
enforcement: mandatory
- clause_id: SOX-802
title: "Criminal Penalties for Altering Documents"
citation: "Pub. L. 107-204 § 802; 18 USC § 1519"
full_ref: "SOX § 802 — 18 U.S.C. § 1519 destruction/alteration of records; 20-year imprisonment"
lambda_axes:
- axis: Λ7
label: Auditability
weight: 1.0
enforcement: mandatory
rationale: >
Receipt chain immutability is a legal requirement; SHA3-256 Merkle
DAG nodes must be append-only with cryptographic proof of non-alteration.
- axis: Λ6
label: Security
weight: 0.95
enforcement: mandatory
compliance_thresholds:
minimum_lambda_coverage: 7
mandatory_axes: [Λ2, Λ7, Λ9]
receipt_retention_days: 2555 # 7 years SOX record retention
max_decision_latency_ms: 500
model_validation_cycle_days: 365
receipt_chain:
algorithm: SHA3-256
chaining: merkle_dag
quorum: 2-of-3
nodes: [primary, secondary, regulatory-archive]