Definition
A phishing attack that tricks a user into granting an AI agent permission through an identity provider, planting a permanent, hard-to-revoke agent identity inside the company's systems. It abuses the 'sign in with X' consent flow that agents use. The planted agent can quietly live inside the tenant and act on the attacker's behalf.
Why it matters
It lets an outsider implant a credentialed agent that bypasses normal device and account controls, and it is hard to detect or undo.