Here’s How OpenAI Says It Used AI to Find a Spy Inside Its Company

OpenAI says it used its own AI models to root out what sounds like some kind of alleged corporate spy. They say they handed them over to the FBI, who made an arrest on October 9.

OpenAI told Gizmodo the employee had been working in product at OpenAI since July of last year.

According to OpenAI, it performs regular security sweeps, and during one such sweep of a conference room on September 30, its technical surveillance countermeasures found something suspicious.

A statement provided to Gizmodo refers to the apparent bug as a “sophisticated recording device.”

The device itself monitored audio, and could be used to transfer data, according to OpenAI. The company says the device used encryption, and was designed for stealth.

The device apparently stored data, and according to OpenAI, the company’s investigators were able to feed forensic images of that storage, and of some kind of data changes, into its own AI models, and that when combined with analysis from its investigators, this AI-enhanced process apparently led the company to a suspect.

According to OpenAI, there is evidence suggesting that this same employee was planning to install more bugs.

Protecting AI technology from state-level actors is considered a high priority for the U.S. Treasury Secretary Scott Bessent has, for instance, floated the idea of sanctioning China over alleged distillation, a way of copying the functionality of AI models that requires minimal secrecy, let alone any sort of bugging of an OpenAI office.