SesameOp Backdoor Uses OpenAI API for Covert C2

A new backdoor dubbed SesameOp uses an OpenAI API for covert command-and-control communications to manage malicious activity within a compromised environment, illustrating how threat actors can abuse generative AI services. Researchers from Microsoft’s Detection and Response Team (DART) discovered SesameOp in July after responding to a security incident in which attackers remained undetected for months, according to Microsoft Incident Response.

The backdoor includes a component that uses the OpenAI Assistants API as a storage or relay mechanism to fetch commands, with the malware then executing them, and the attackers employed techniques such as payload compression and layered encryption to hide C2 communications. The infection chain comprises a loader (Netapi64.dll) and a NET-based backdoor (OpenAIAgent.Netapi64) that leverages OpenAI as a C2 channel, as described in the post.

Mitigations include auditing firewalls and logs, employing local, network and perimeter protections, and configuring proxies to limit access to services, with OpenAI noting the API key would be disabled and that the account had limited API calls.