← All services
Practice 02
AI Model Risk Management & Regulatory Compliance
SR 11-7-aligned model risk frameworks that hold up under examination — adapted for the realities of generative AI.
SR 11-7 was written for traditional models. Applying it to LLMs and agentic systems is not optional in regulated banks, but the literal interpretation falls apart on probabilistic, prompt-conditioned systems. We bring a working framework — extended for non-determinism, prompt drift, and tool use — and the documentation templates examiners expect to see.
Outcomes
What you walk away with
- Model inventory and tiering schema that includes GenAI and agentic systems
- Validation playbooks for LLM-specific risks: hallucination, jailbreak, prompt injection, drift, bias
- Evaluation harness with quantitative metrics and structured human review
- Examiner-ready model documentation, including assumptions, limitations, and ongoing monitoring plans
Engagement shapes
How we work together
- Framework gap assessment against current MRM policy
- Validation of one or more priority models against the extended framework
- Standing model-risk advisor retainer for ongoing reviews
Audience
Who this is for
Banks subject to Federal Reserve and OCC supervision
Mortgage originators and servicers under Fannie Mae / Freddie Mac frameworks
Credit scoring and consumer-finance analytics firms
Risk and validation teams in pharma, healthcare, and energy
Next step
Ready to scope a ai model risk & compliance engagement?
A 30-minute call gets us to a yes, no, or a clear next step.
Other practices