Source: Agencies Overhaul Model Risk Management Guidance for Banks (Orrick)

Law-firm analysis of the 2026 interagency MRM guidance: narrower model definition, materiality-based tiering, a lighter validation description, non-enforceability, and a carve-out for generative and agentic AI.

Contents

Summary

Compares the April 2026 OCC/Fed/FDIC guidance with SR 11-7 point by point.

Key claims

  • Also rescinds OCC Bulletin 2011-12, FDIC FIL-22-2017, OCC 1997-24 on credit scoring, and the Comptroller’s Handbook MRM booklet. The applicability threshold rises from $1B to $30B in assets.
  • A “model” now requires complexity. Simple arithmetic, spreadsheets and deterministic rules are carved out.
  • A footnote excludes generative and agentic AI as novel and fast-changing. Banks should govern them through broader risk management, and the agencies plan a request for information on AI.
  • Model risk = inherent risk (complexity, assumptions, data quality) combined with materiality. Low-risk, immaterial models may get lighter oversight.
  • Conceptual soundness, outcomes analysis and ongoing monitoring remain. Effective challenge is redefined around expertise, independence and standing to effect change.
  • Third-party models: the principles still apply even when the vendor’s model cannot be fully validated, and the bank’s customizations must be documented and evaluated.
  • The guidance is not enforceable in itself. Unsafe or unsound practices can still be pursued.

Feeds

AI governance and compliance

Suggest changes →