Runtime Policy

A runtime policy is a governance rule that is evaluated and enforced while an AI agent is actively executing, rather than at deployment time. Runtime policies respond to real-time context — the data being accessed, the current risk score, the identity making the request — and can block, throttle, log, or escalate actions without requiring a redeployment.