NIST AI Risk Management Framework

A practical framework for governing, mapping, measuring, and managing AI risk.

The NIST AI Risk Management Framework gives organizations a common structure for thinking about AI risk. Its four functions (Govern, Map, Measure, and Manage) connect policy and responsibility to the context in which a system is actually used. It encourages teams to define intended use, affected people, likely harms, measurement limits, and ongoing oversight.

We use the framework as a way to keep evaluation close to deployment. A tidy benchmark cannot describe every source, language, time pressure, or consequence a live system will encounter. The framework is voluntary and broad by design; teams still need to translate it into concrete review rules, ownership, tests, and correction paths.