An OpenAI engineer has been arrested and charged with using or attempting to use a device to intercept oral communications after the company discovered a concealed recording device in an executive conference room. OpenAI said it fired the employee and referred the incident to the FBI, whose investigation remains ongoing.
The employee was identified as Abdulrahman Alfozan, a 35-year-old Saudi national. He was arrested Friday and made his first appearance in federal court Tuesday, remaining in custody during that period. The charge is an allegation, not a finding of guilt.
Federal prosecutors said the device was connected to a television’s USB port in a room used for discussions involving sensitive technical, commercial and strategic information. It was covered in toilet paper and installed on or around September 28. Its contents were encrypted, leaving unresolved in the provided account what communications, if any, it had captured.
Alfozan did not have access to the executive conference room, prosecutors said. They alleged that surveillance footage showed him near the room on several dates before the device was discovered. Footage also allegedly showed him appearing to look behind the television where the recorder was installed. Those observations form part of the government’s account of his connection to the device.
OpenAI security personnel also found a bag that Alfozan had left in another room at the office, prosecutors said. The bag contained his wallet, identification and a second recording device. The account does not establish whether that second device had been used to record conversations.
OpenAI said its own AI models assisted its examination of the concealed hardware and its origins, helping the company identify an employee it concluded had placed the device several days earlier. The company did not describe the models, the analysis they performed or how their output was checked. Its statement therefore provides limited detail about the technical role AI played in the inquiry, distinct from the surveillance footage and physical evidence described by prosecutors.
The company also said it had found no indications that the employee improperly accessed or removed model weights, source code or customer data. That statement distinguishes the suspected interception of spoken conversations from the theft of digital assets. It does not establish whether confidential discussions were recorded or whether anyone obtained information from the device.
The case concerns a form of physical surveillance rather than a disclosed intrusion into OpenAI’s computer systems. Conference-room conversations can contain sensitive information even when participants do not open files or transmit data electronically, making access to rooms and equipment a separate security concern from controls on software repositories and model files.
OpenAI said it would continue assisting the FBI. The investigation has not publicly resolved what the device recorded, whether its contents were shared or what motivated its alleged placement.
Sources: The Verge (AI)