Glossary
Policy as code
Writing the rules an AI system runs under, who may call what, with which data, with what limits, as versioned configuration that is reviewed like code and enforced by the platform.
The alternative is rules that live in people's heads and in chat prompts, which an agent can be talked out of. In a platform the policy configuration sits beside the agent and model configuration in its control plane, every change is a reviewed commit, and the policy enforcement point reads it at run time.