OpenAI Engineer Charged With Bugging the Company's Own Executive Conference Room

Federal prosecutors in San Francisco have charged an OpenAI product-team engineer with secretly recording conversations in one of the company's own executive conference rooms. Abdulrahman Alfozan, a 35-year-old Saudi national who joined OpenAI in July 2025, is accused of planting a hidden device behind a conference-room television, concealed in a wad of toilet paper and wired into the TV's own USB port. He was fired and arrested on Oct. 9, made his first federal court appearance on Tuesday, and remains in custody. OpenAI says it has found no indication he took model weights, source code, or customer data -- but the case is still at the allegation stage, and neither the criminal complaint nor the company has said what, if anything, the device actually captured before it was found.

The charge itself is narrow: using a device to intercept oral communications, a single federal count that carries up to 5 years in prison, a $250,000 fine, and 3 years of supervised release if he's convicted. Prosecutors describe the alleged preparation as methodical rather than impulsive. In the days before the device went in, an account investigators tie to Alfozan's household allegedly used an AI assistant to ask about the building's floor plan, conference-room capacities, and where leadership and partner meetings were held -- then, closer to the install, about the recorder itself, including its MAC address. OpenAI dates the installation to around Sept. 28; staff found the device two days later, on Sept. 30.

What happened next is the part OpenAI's own account treats as a vindication of its tools. The company's technical surveillance countermeasures team found the recorder during a routine sweep, and OpenAI says it then used its own AI models -- not an outside forensics firm -- to identify the device and trace it to Alfozan before handing the case to the FBI. Investigators separately used those same models to analyze forensic images pulled from the device's own storage, according to OpenAI's account -- the company's tools turned on the hardware that had been aimed at it. OpenAI had also placed a camera behind the same television, and on Oct. 5 that camera allegedly recorded Alfozan returning to the room and looking behind the TV where the device sat. The same day, a second, similar recording device turned up in an unattended bag in a different room. OpenAI describes the hardware itself as carrying advanced encryption, audio-monitoring capability, and a data-transfer mechanism "designed to evade monitoring" -- language that, if accurate, points to purpose-built covert-collection hardware rather than an off-the-shelf recorder.

OpenAI's response between discovery and arrest was gradual rather than immediate: it says it began restricting Alfozan's access on Sept. 30, the day the first device was found, then suspended his physical and network access, placed him on administrative leave, and secured his company equipment, before the FBI made the arrest more than a week later. That sequencing matters for what it implies about the investigation's pace -- OpenAI was managing a known internal risk for over a week before federal charges existed.

U.S. Attorney Craig Missakian called the case "a matter of national security." FBI Director Kash Patel went further, telling reporters the scheme aimed to "capture sensitive communications about the company's super intelligence technology" and describing protection of American AI labs as among the bureau's "highest counterintelligence priorities." Those are characterizations, not findings a court has made: the complaint describes what prosecutors allege Alfozan did, not why he allegedly did it. His motive is not stated anywhere in the public record, and Alfozan and his attorney could not be reached for comment. His career history -- five years as a senior systems engineer at Meta, then roughly three as a founding engineer at the startup Nash, then OpenAI since mid-2025 -- doesn't itself point toward foreign direction any more than it rules it out.

OpenAI, for its part, said it treats protecting its technology, data, and people "as a matter of national importance," and thanked the agents and prosecutors who "moved quickly to make an arrest." The company has not said whether it is changing physical-security procedures company-wide, beyond the specific access restrictions it applied to Alfozan himself. (The case surfaces in a month when AI labs' internal-risk posture has been unusually visible in public: Microsoft has disclosed an internal "emergency brake" policy for AI-related insider risk, and Anthropic has disclosed that one of its own models filed a false tip with Philadelphia police. None of those events are related to this one -- they just happen to share a month, and a theme: what AI labs are willing to say about what can go wrong inside their own walls.)

OpenAI's public statements describe what the company is confident did not happen rather than what did: no sign of access to model weights, source code, or customer data, and no compromise of product safety that it has found so far. That is a lower bar than "nothing was recorded," and it leaves the central question -- what, if anything, Alfozan's device actually captured before it was found -- open heading into his detention hearing.