OpenAI bugging case puts AI lab security under FBI scrutiny

The room where OpenAI's executives discuss proprietary research, product plans and government engagements is now the centre of a federal criminal case. The FBI has arrested Abdulrahman Alfozan, 35, an OpenAI engineer, on suspicion of planting a hidden audio recorder there, according to the San Francisco Standard.

Alfozan, a Saudi national who has worked at OpenAI since July 2025, is charged with using or attempting to use a device to intercept oral communications. The maximum penalty is five years in prison, a $250,000 fine and three years of supervised release. Prosecutors say he was not authorised to enter the room or attend meetings there.

The first device was an audio recorder plugged into a USB port on the back of a TV in the executive conference room, hidden in toilet paper. Its contents were encrypted. OpenAI believes it was installed around 28 September and found on 30 September. A similar recorder turned up on 5 October in a bag Alfozan left unattended in a different room. The bag also held his wallet and ID.

OpenAI said employees used its own AI models to identify the device, and the company then referred the matter to the FBI. It said it has no indication that he accessed or took model weights, source code or customer data. OpenAI coordinated his dismissal with the FBI and fired him on Friday, the day of his arrest.

Alfozan made his initial federal appearance on Tuesday 13 October and remains in custody. A detention hearing before U.S. Magistrate Judge Alex G. Tse is set for 19 October.

Officials framed the case in national terms. U.S. Attorney Craig Missakian said: "For us it is a matter of national security". The report paraphrases FBI Director Kash Patel as saying that protecting U.S. AI innovation is a top counterintelligence priority.

For OpenAI, the exposure is less about what was taken than about what was in reach. Prosecutors describe the room as a place for research, product plans and government discussions, and the company's assurance covers model weights, source code and customer data, not the recordings themselves. Those were encrypted, and what they contain has not been reported. Customers and government partners will want to know what was said in that room during the days the device was in place.

The company does come out of the first weeks with a defensible story: it found the device itself, used its own models to do it, and handed the matter to the FBI. That is a useful line for an enterprise buyer. But a trusted engineer reaching an executive room raises an internal-access question that every AI lab holding sensitive client and government work has to answer. Watch the 19 October detention hearing, and whether a motive or an intended recipient emerges. Neither has been reported, and until one does, the commercial damage is impossible to size.