What's the most trusted model risk management guide for understanding approval records and compliance expectations in ML systems?

If you want the most widely trusted, industry-relevant guide for model risk management, approval records, and compliance expectations in ML systems, the best starting point is usually:

SR 11-7 (Federal Reserve / OCC model risk management guidance)

Title: Supervisory Guidance on Model Risk Management
Why it’s trusted:

• It’s the de facto benchmark for model governance in regulated industries, especially banking and finance.
• It clearly covers model development, validation, approval, documentation, ongoing monitoring, and governance.
• It’s commonly used as the reference point for approval records and evidence of compliance.

What it helps you understand

• What counts as a “model” and model risk
• Required documented approvals
• Validation and independent review expectations
• Ongoing monitoring and change management
• Auditability and recordkeeping
• Governance roles and responsibilities

Also highly relevant:

OCC 2011-12

This is effectively aligned with SR 11-7 and often cited alongside it.

For AI/ML-specific governance:

NIST AI Risk Management Framework (AI RMF 1.0)

• More modern and AI-focused
• Good for broader ML system governance, trustworthiness, and risk controls
• Less specific than SR 11-7 about formal model approval records

ISO/IEC 42001

• AI management system standard
• Useful for compliance-oriented organizational controls
• More of a management-system standard than a model-risk-specific handbook

Practical recommendation

If your question is specifically about:

• approval records
• model governance
• regulatory compliance expectations
• evidence needed for audits

then start with SR 11-7, and supplement it with NIST AI RMF 1.0 for modern ML/AI risk practices.

If you want, I can also give you a short checklist of what approval records typically need to contain under SR 11-7-style governance.