Chatbot prompts are evidence in the OpenAI bugging case

Your chatbot history is a record. In the case of an OpenAI engineer charged with bugging an executive conference room, prosecutors are using one.

The Justice Department said on Oct. 13 that Abdulrahman Alfozan, 35, a Saudi national, was charged with using a device to intercept oral communications. He was arrested Oct. 9. These are allegations in a criminal complaint, and he has not been convicted.

A camera, a meeting-room search and a chatbot account

According to the Justice Department, Alfozan had worked as an engineer at OpenAI since July 2025. On or about Sept. 30, OpenAI staff found a hidden audio recording device in a USB port on the back of a television in an executive conference room in San Francisco. OpenAI determined it had been installed on or about Sept. 28, hidden in toilet paper, with encrypted contents.

The complaint cites three kinds of evidence, per the department's release.

Security camera footage showed him near the room on several dates, though he had no authority to be in its meetings. A productivity tool logged his questions about building addresses, meeting rooms, room calendars and "leadership and exec" locations.

And a chatbot account. Prosecutors say an account tied to his spouse, which shared identifiers with his own, sent prompts from Sept. 24 to Sept. 30 about building, installing, testing, obfuscating and remotely accessing a recording device. That included inquiries about "build[ing] a continuous audio recorder" and "encrypt[ing] the SD card."

The same account asked about the device's custom header and MAC address on Sept. 27 and 28, the complaint alleges. The release does not name the chatbot.

On or about Oct. 5, the complaint says, security found a second, similar device in a bag he left in another room.

He faces up to five years in prison and a $250,000 fine if convicted.

OpenAI said in a statement, reported by TechCrunch, that it has "no indications of this individual inappropriately accessing or exfiltrating model weights," source code or customer data. TechCrunch reports he was fired.

The threat to AI labs is down the hall

The usual fear at an AI lab is an outside hacker stealing model weights. This case, as alleged, is the cheaper attack: an employee with a badge and a small device behind a TV.

Washington is treating it as national security. FBI Director Kash Patel said Alfozan allegedly hid the device "to secretly capture sensitive communications about the company's super intelligence technology." The department says the prosecution supports a presidential memorandum, NSPM-11, on American superiority in superintelligence. Expect that framing to show up in how labs screen and watch their staff.

Your prompts are a paper trail

Two lessons, and neither needs you to work at OpenAI.

First, prompts are evidence. Whatever you type into a chatbot sits in an account log that a court can reach. This complaint ties a spouse's account to a suspect through shared identifiers, prosecutors allege. Write your prompts as if someone may read them later, because someone may.

Second, insider risk is the boring threat that works. If your company handles anything a rival or a government wants, notice what caught this: a camera on the conference room and a log of who asked about the executive calendar. No firewall was involved.

Watch the indictment and the policy that follows

A complaint is the first step. Watch for an indictment and for whether prosecutors allege anyone directed him. Watch, too, for labs tightening physical security and employee monitoring in the name of NSPM-11.

Assume the log exists. Act like it.