Microsoft has launched the public preview of Entra Agent ID, extending its identity-and-access governance model to non-human AI agents and promising to rein in “shadow AI” workflows now spreading across large enterprises.
Entra already processes about 8 billion authentications a day and has pilots managing 10,000 AI agents inside a single organization—volume that underscores how quickly agentic systems are moving from experiment to production. With Agent ID, each AI agent can be registered, sponsored, governed through lifecycle policies, and bound to conditional access and identity protection controls that were previously focused on human users, according to a report by VirtualResource (PDF).
The move effectively shifts the identity race into AI territory. AWS recently introduced Bedrock AgentCore Identity for securing AI agents, while startups are pushing decentralized “agent identity” frameworks built on DIDs and verifiable credentials, reports Sp-Edge. Traditional workload-identity tools from vendors such as Okta, Ping, and other Gartner-tracked “workload identity management” providers now face a world where agent identity volume may eclipse container and microservice identities.
The launch puts Microsoft a step ahead on centralized AI-identity governance—but also raises a question: will customers accept a tightly coupled model that deepens dependence on Entra, or will multi-cloud shops seek more neutral “agent identity” control planes?
Stay on point and visit the latest SecPo Marketplace headlines.
