Observed incident

Amazon decoys captured AI agent logs describing credential theft and attempted identity fraud

Reported

Published here

What happened

Between October 3 and November 5, Amazon’s MadPot decoys received requests containing AI agent histories. The logs described efforts to steal keys to pay for AI use and to pass identity checks at two cryptocurrency exchanges.

Parties and reporting

Where this happened

Victim location not disclosed. Real organizations or systems were affected or targeted, but the sources do not say where. Explore the world map.

AI tools and systems: ARTEX, Amazon MadPot, AI inference services

Outcome

Amazon notified potentially affected customers where appropriate. For one identity verification provider, the agent logs recorded rejection of every deepfake attempt.

Operating lesson

Review unexpected inference charges alongside access to API keys. Use independent verification when investigating suspicious identity checks.

Evidence and limits

Amazon disclosed the findings on November 13 and assessed ARTEX use with moderate confidence. Its sensors observed incoming requests. Activity and outcomes on external systems were reported by the agents and remain independently unverified.

Sources