Policies and assignment

An Actera policy is a reusable set of authorization rules for a target AI agent action. Publication creates an immutable version that can be evaluated consistently and retained in decision evidence.

Publication and assignment are separate:

• Publication makes a reviewed version available.
• Assignment connects that policy to one or more agents.

This separation supports composition. Two agents may reuse a shared policy, while a refund-specific policy can apply only to the refund agent. All applicable assigned policies contribute to the final result.

Rules produce ALLOW, REQUIRE_APPROVAL, or BLOCK. Actera resolves concurrent matches using the precedence shown in How Actera works. Follow the policy configuration guide to test, publish, and assign a version.