As AI agents and agentic architectures become widely adopted in production environments, their growth poses new and complex identity-centric cybersecurity challenges. AI agents will need to be authenticated to critical systems, they will need strict access controls and they must be restricted only to performing their intended functions.