Compliance for AI systems is the practice of ensuring AI behavior, data handling, access, monitoring, and documentation meet legal, regulatory, contractual, and organizational requirements. Requirements vary by industry, geography, and use case.
From an engineering perspective, compliance needs instrumentation, access control, retention policies, audit logs, evaluation records, human review workflows, and evidence that policies are enforced. Compliance cannot be added only as copy in a prompt.