Glossary
Updated
A formal policy document, enforced via technical controls (role-based access, policy-as-code engines), that delineates who may use an AI system, for what tasks, and under what conditions. It lists prohibited use cases (e.g., mass surveillance, deceptive advertising), required approvals for sensitive operations, and monitoring protocols. Governance automates policy checks at runtime and regularly reviews policies to capture new risks or regulatory changes.
Example
A research lab’s Sanctioned Use Policy requires any team wanting to use its facial-recognition API to obtain ethical-committee approval, sign end-user–license agreements forbidding surveillance, and log every invocation. A policy-engine gateway enforces these rules automatically at the API layer.