logo

Hackers Could Abuse Microsoft Entra AI Agents to Hijack Enterprise Accounts Without Stealing Passwords

ID: 271a1e00-750b-54b9-9713-58c228cb33c9

STIX ID: report--271a1e00-750b-54b9-9713-58c228cb33c9

Feed Name: cybersecurityNews.com

Threat Score
65/100

Date Published: 2026-11-11

Date Updated: 2026-11-11

Author: Tushar Subhra Dutta

...
...

Researchers (Proofpoint) describe a proof-of-concept 'BluePhishing' attack that leverages Microsoft Entra Agent ID and OAuth-style consent to get victims to approve a malicious agent blueprint; once consented, the blueprint principal receives AgentIdentity.CreateAsManager (reported as non-revocable), allowing attackers to create agent identities with broad Graph permissions (e.g., User.Read.All, Group.Read.All). The report details how attackers build blueprints in their tenant, obtain tenant and user identifiers, and use tokens to run arbitrary Graph requests, lists IoCs (permissions, signInAudience setting, audit events, blueprint ID), and recommends mitigations such as user training, limiting agent consent, permission classifications, least privilege, Conditional Access and audit log monitoring.

Your team is not currently subscribed to this feed. You must subscribe to it in order to see this post.