Microsoft Entra ID vulnerability allowed global admin impersonation

As outlined in HackRead, cybersecurity researchers have uncovered a significant vulnerability within Microsoft Entra Agent ID, a platform designed to manage AI identities. This flaw could allow attackers to impersonate global administrators within an organization's network.

The vulnerability, discovered by Silverfort researchers, resided in the Agent ID Administrator role. While intended to manage AI agent identities and related objects, the role's permissions were too broad, allowing administrators to modify nearly any Application Service Principal within a Microsoft Entra tenant. Attackers could exploit this by adding themselves as an owner to a high-privilege Service Principal, effectively hijacking its identity. This allows them to gain full control over the network, as demonstrated by a successful hijack of a global administrator account.

The implications are widespread, as approximately 99% of business networks utilize privileged Service Principals. Microsoft has since released a fix, rolling out a patch on April 9, 2026, that restricts the Agent ID Administrator role from managing owners of non-agent Service Principals. Organizations are advised to review their audit logs for suspicious changes to account ownership or the creation of new secrets on sensitive accounts.

Source: HackRead