Microsoft Warns a Key OpenAI API Is Being Exploited to Launch Cyberattacks

Why It Matters

The abuse demonstrates how legitimate AI services can be weaponized for stealthy cyber‑espionage, raising concerns for enterprises that integrate such APIs without robust monitoring. It also highlights the need for tighter oversight of third‑party cloud services used as C2 channels.

Summary

Microsoft’s incident‑response team has identified a new malware family, SesameOp, that abuses OpenAI’s Assistants API as a covert command‑and‑control channel. The backdoor stores and retrieves encrypted commands and exfiltrated data via the API, enabling long‑term persistence for espionage‑type operations. Microsoft stresses the abuse leverages a legitimate feature of the API, not a vulnerability, and notes the Assistants API is slated for deprecation in August 2026. Researchers advise organizations to audit firewall logs, enable tamper protection, and monitor for unauthorized external‑service connections.