Version 1.1 clarified more than it changed — but three sections rewrote how auditors sample controls.
The 1.1 revision to the NIST AI Risk Management Framework arrived quietly, but if you read it as an auditor rather than a policy author, three sections stand out.
- The Map function got sharper
‘Map 3.5’ now explicitly names third-party AI components as in-scope. If your firm consumes a foundation model through an API, that dependency is part of your risk surface — and your evidence trail. - Measure introduced continuous monitoring
The framework now treats measurement as an ongoing activity, not a one-time validation. This is the change that most affects audit sampling: you can no longer test the model once and call it governed.
