A practical mapping between the major AI governance frameworks and the runtime primitives that enforce them: identity, policy, evidence, escalation.