Glossary / SR 11-7

Definition

What is SR 11-7?

Model risk management guidance

SR 11-7 is the U.S. Federal Reserve and OCC supervisory guidance on model risk management. It requires institutions to manage the risk of model errors through three pillars — conceptual soundness, ongoing monitoring, and outcomes analysis — backed by effective challenge. It applies to AI and LLM models, not only traditional statistical models.

Because a generative or agentic AI system takes inputs and produces outputs that inform business decisions, it meets the SR 11-7 definition of a model and falls inside an existing model risk management program — with the same examiner expectations.

Trinitite satisfies the three pillars with signed evidence: content-Merkle model documentation gated by effective challenge (conceptual soundness), PSI drift detection on a streamed event feed (ongoing monitoring), and deterministic replay for reproducible outcomes back-testing. It crosswalks to Fed SR 11-7, OCC 2011-12, and OSFI E-23.