Who Let an Unauthorized Engineer Into OpenAI's Conference Room?

OpenAI caught the spy with its own AI. It still has to explain why the spy got inside the room in the first place.

On September 30, OpenAI personnel found a thumb-sized device plugged into a USB port on the back of a TV in an executive conference room at its San Francisco headquarters. It had a microphone, could connect to a camera, and its contents were encrypted. Federal prosecutors charged Abdulrahman Alfozan, 35, a Saudi national who joined the company in July 2025, with using a device to intercept oral communications. If convicted, he faces up to five years in prison.

Here is the part everyone is skipping: CCTV footage showed Alfozan appeared near the executive conference room on multiple dates before the device's discovery, although he had no authority to access it. Not once. Multiple dates. In a room used routinely for discussions involving highly confidential technical, commercial and strategic information, including proprietary research, product plans, and government engagements. That is the room where OpenAI talks to the U.S. government. A product engineer with no clearance walked past whatever door control exists and nobody pulled the footage until after the device turned up.

The Digital Trail Was There the Whole Time

The paper trail is damning in hindsight. Prior to September 28, Alfozan used a productivity tool to search OpenAI's building addresses, meeting rooms, room capacities and calendars, and locations of "leadership and exec" meetings. Separately, a chatbot account tied to his spouse, which shared identifiers with his own, submitted prompts between September 24 and 30 about building, installing, obscuring, and remotely accessing a recording device, including queries about how to "build a continuous audio recorder" and "encrypt the SD card." On September 27 and 28, the account asked about the specific device later found, including its unique MAC address.

None of that triggered an alert before the device was physically discovered. The system that caught him was retrospective, not preventive.

OpenAI's statement is where the spin lives. The company says it used its internal AI models to identify the person responsible and trace the device's origins. Translation: our model is a great forensic tool after the fact. It found the fingerprints. It did not stop the break-in. A second, similar device turned up in Alfozan's personal bag, meaning the plan wasn't a one-off.

Give OpenAI credit for the cleanup. The company gradually restricted his access between September 30 and October 9 and maintains it has no indications he inappropriately accessed or exfiltrated model weights, source code, or customer data. FBI Director Kash Patel called this a counterintelligence priority. That is the right posture for the Bureau.

The verdict: an insider threat that walked past every lock except the one that photographs.

The company's job now is to answer a narrower question its own AI didn't ask: who controlled that room's calendar, who had badge access, and how many weeks of unauthorized entries get logged before a human looks at the tape.

What to watch: the detention hearing set for October 19, whether the FBI complaint ties Alfozan to anyone beyond his spouse's account, and whether OpenAI's board discloses an actual access-control remediation plan instead of another AI-caught-it press release.