Agentic AI risk management is the security and risk program around autonomous AI agents. It defines who owns each agent, how scopes are reviewed, how approvals are routed, how credentials are rotated, and how the audit trail is preserved.