Glossary term
Glossary term
Security
System that evaluates rules and permissions before actions are allowed.
Open Policy Agent (OPA) is integrated by enterprise AI platform teams to evaluate whether an agent action (e.g., 'delete customer record') is permitted based on the agent's role, user role, and record sensitivity.
Microsoft Purview acts as a policy engine for Microsoft 365 Copilot - defining which SharePoint sites, Teams channels, and Exchange mailboxes the agent can access based on information-protection labels.
Aporia's ML monitoring platform acts as a policy engine for LLM agents - evaluating each model input/output pair against business rules, safety policies, and fairness constraints before serving.