Microsoft Fixes Entra ID Flaw to Prevent Abuse of AI Agent Administration Privileges

Microsoft Entra ID is a cloud platform that enterprises use to manage user, application and service identities. Its Agent ID Administrator role is responsible for configuring AI agent identities. If those privileges were abused, an attacker could circumvent existing restrictions and take control of other service principals within an organization, increasing the risks of data breaches and lateral movement.

Cybersecurity company Silverfort discovered and reported the privilege flaw, saying attackers could use an AI agent identity to take control of any service principal. Microsoft subsequently revised the authorization mechanism for the Agent ID Administrator role and completed remediation across its global Entra ID cloud environment in April. It has not disclosed the number of affected customers, any financial losses or any cases of active exploitation.