SesameOp Backdoor Explained: What You Need to Know About the OpenAI API Exploitation

In July 2025, Microsoft uncovered a new backdoor named SesameOp, which marks a notable shift in cyber-espionage tactics by exploiting the OpenAI Assistants API to communicate with and exfiltrate data from compromised devices. According to Microsoft’s findings, SesameOp is a covert backdoor that abuses the OpenAI Assistants API to issue commands, execute payloads, and relay results between attackers and infected hosts, maintaining long-term persistence through legitimate cloud infrastructure.

The infection chain centres on a loader named Netapi64.dll and a .NET-based backdoor called OpenAIAgent.Netapi64, with obfuscation applied via Eazfuscator[.]NET to hinder reverse engineering. SesameOp uses the OpenAI Assistants API as a command-and-control channel, retrieving commands, decrypting and executing payloads, and then returning results as encrypted messages, all designed to blend with normal API traffic.

Microsoft’s investigation found the campaign sought espionage and sustained presence rather than destruction, aided by multiple compromised Visual Studio utilities and careful operational security. OpenAI and Microsoft subsequently disabled the malicious API key and account, according to Microsoft’s official blog.

For defenders, the report emphasises the need for behavioural detection and API auditing alongside traditional network controls, and notes that further technical details and mitigations are available on the Microsoft blog.