Tool

Which AI rules apply to this capability?

Answer six questions about the organization and the decision. The list narrows to the rules that apply, with the evidence each asks for.

Sector
Bank details
Where you operate
The capability

  • SR 26-2 / OCC Bulletin 2026-13, Revised Guidance on Model Risk Management (17 April 2026) Nonbinding guidance

    What it asks: Model risk management scaled to materiality. Generative and agentic AI are outside its scope, so name the internal policy that governs them.

    Evidence: System record, evaluation and release decision under a named internal policy.

    Source

  • OCC Bulletin 2026-54, Super Intelligence: Supervisory Framework (4 November 2026) Examiner framework

    What it asks: What examiners will look at for generative and agentic AI: sixteen components in six pillars (purpose, risk, control, oversight, resilience, assurance). Sets no new expectations for banks.

    Evidence: Tag each artifact in the evidence package with the pillar it answers.

    Source

  • Interagency Guidance on Third-Party Relationships (June 2023); replacement proposed 11 September 2026 In force; revision proposed

    What it asks: Risk-based due diligence, contract terms, monitoring and exit for model and AI providers. The 2023 guidance applies until the proposal is final.

    Evidence: Supplier records, tested integration and an exit plan.

    Source

  • ECOA and Regulation B, adverse action notices (12 CFR 1002.9) Regulation

    What it asks: Specific principal reasons for a credit denial or other adverse action, however complex the model.

    Evidence: Reasons traceable to the request trace, with tested explanations.

    Source

  • NYDFS Part 500, with the October 2024 industry letter on AI Regulation and guidance

    What it asks: AI risks addressed within existing risk assessment, third-party, access, training, monitoring and data controls.

    Evidence: Threat model, access and revocation tests, monitoring of unusual queries.

    Source

  • NAIC Model Bulletin on the Use of AI Systems by Insurers (December 2023) Model bulletin; check each state

    What it asks: A written AI systems program: governance, risk management, internal controls, third-party oversight and testing for adverse consumer outcomes, where your states have adopted it.

    Evidence: Applicability record, supplier records, evaluation results and release decisions as program evidence.

    Source

  • NYDFS Insurance Circular Letter No. 7 (July 2024) Circular letter

    What it asks: Governance, risk management and fairness testing for AI and external consumer data in underwriting and pricing.

    Evidence: Data source approvals, and evaluation sets and release criteria with fairness tests.

    Source

  • Colorado Regulation 10-1-1 (life insurers; private auto and health benefit plans from 15 October 2025) Regulation

    What it asks: A risk-based governance framework, an inventory of external data, algorithms and models, board and vendor oversight, and annual compliance reports.

    Evidence: System record and inventory, owner and board reporting, supplier records.

    Source

  • Colorado SB 26-189, automated decision-making (signed 14 May 2026) Enacted; effective 1 January 2027

    What it asks: Duties for developers and deployers of automated decision-making that materially influences consequential decisions. Due to apply from 1 January 2027; confirm current enforcement status with counsel.

    Evidence: Applicability record and notices; track the litigation.

    Source

  • EU AI Act, as amended by the Digital Omnibus (Regulation (EU) 2026/1744) Regulation; staged dates

    What it asks: Annex III high-risk obligations from 2 December 2027, including creditworthiness of natural persons and life and health insurance risk and pricing; Article 50(2) marking of generated content from 2 December 2026; AI literacy and prohibitions already apply.

    Evidence: Classification record, technical documentation, human oversight design and post-market monitoring.

    Source

  • DORA, Regulation (EU) 2022/2554 (applies from 17 January 2025) Regulation

    What it asks: ICT third-party risk, a register of information, contract terms and major ICT incident reporting.

    Evidence: Supplier records, exit plans and the incident runbook.

    Source

  • PRA SS1/23, Model Risk Management Principles (effective 17 May 2024) Supervisory statement

    What it asks: Model inventory, tiering, independent validation and governance, including AI-based models.

    Evidence: System record, version evidence and independent review.

    Source

  • Financial Stability Board, sound practices for responsible AI adoption (final, 15 October 2026) Voluntary

    What it asks: Twelve voluntary practices: organization-wide governance, lifecycle controls, clear triggers for human approval of agent actions and tested ways to stop them.

    Evidence: Unit of approval, action limits with approval points, and the switch-off test.

    Source

  • NIST AI RMF 1.0 and the Generative AI Profile (NIST AI 600-1) Voluntary

    What it asks: A voluntary structure for risk work: Govern, Map, Measure, Manage. Under revision in 2026; record the version you use.

    Evidence: Risk register and the mapping of controls to the framework.

    Source

Checked 15 November 2026. This is a delivery aid, not legal advice; your legal and compliance owners decide which rules apply.

Companion to AI Governance Evidence. Everything you enter stays in this browser.

Next step

Map the evidence once, for every reviewer.

We build the evidence package around the deployed capability, tagged to each rule that applies.