SR 26-2 validation pack
Evidence and method for your model-risk team, shared on request under NDA.
SR 26-2 replaced SR 11-7 in April 2026. It excludes deterministic rule-based processes from its definition of a model, places generative and agentic AI outside its scope, and still expects sound governance of those tools. This pack helps your team decide what is in scope and test it. It supports your validation and does not replace it: results for your deployment come from your validators, using your data.
What your validators receive.
-
Scope Analysis
Each component mapped against the SR 26-2 definition of a model, with a suggested starting position for your team to accept or challenge.
-
Model Inventory
Starting inventory entries and a template to complete with your owners, tiering, and limitations.
-
Conceptual Soundness
Design choices, assumptions, and known limitations, written for critical review.
-
Reproducibility Test
A procedure for rebuilding the environment from pinned versions and comparing outputs.
-
Outcomes Analysis and Monitoring
Methods for back-testing against your own outcomes, and a proposed monitoring plan.
-
Audit-Trail Evidence
How to capture and verify hash-chained audit records, rule changes, and filing approvals.