Skip to content
NewA calmer, faster ZeroHour
Researchmedium

Proofpoint details BluePhishing attack on Microsoft Entra agent identities

Original title: BluePhishing Attack Abuses Microsoft Entra OAuth Consent to Create Hidden AI Agent Identities

GBHackersby Mayura Kathir ()3 min read
Part of a story covered by 2 sourcesBluePhishing Attack Abuses Microsoft Entra OAuth Consent to Create Hidden AI Agent IdentitiesMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

One admin consent can create persistent agent identities; Proofpoint reports no observed real-world use.

Proofpoint describes BluePhishing, a research proof of concept that extends OAuth consent phishing to Microsoft Entra Agent ID. An attacker-controlled blueprint, after one consent click, can receive AgentIdentity.CreateAsManager and mint agent identities in the victim tenant. Admin consent can grant tenant-wide application permissions; user consent is limited to delegated permissions. Proofpoint says it has not identified in-the-wild activity.

Unconfirmed: Proofpoint presents this as research only and has not identified in-the-wild activity.

Key facts
  • Microsoft documents a limit of 250 agent identities per non-Microsoft blueprint.
  • Proofpoint says AgentIdentity.CreateAsManager cannot be revoked from the principal.
  • Message Center post MC1506733 says default users can create blueprints from November 2026.
  • Inherited permissions are visible only in runtime tokens, not the Entra portal.
  • The same blueprint ID appears in every tenant that consents to it.
Security assessment · AI, from the article

Impact: Persistent agent identities with broad Graph application permissions

ProductAffected
Microsoft Entra Agent ID—
What to do
  1. 1.Train users on consenting to third-party agents and restrict approvals with permission classifications.
  2. 2.Hunt and remove known malicious blueprints using the shared blueprint ID.
  3. 3.Monitor audit logs for Add application and Add service principal events.
Analyze this story with Vani AIImpact, who is affected and what defenders should do, in a new chat

From the original article

3 min read at gbhackers.com

“BluePhishing,” a technique that extends OAuth consent phishing to Microsoft Entra’s AI agent model. The attacker needs only one consent click to create persistent agent identities inside a victim tenant. Proofpoint says this is a research-based proof of concept and it has not identified any in-the-wild activity Microsoft Entra Agent ID extends Entra to AI […]

GBHackers

By Mayura Kathir · gbhackers.com

Read the full article at GBHackers

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.