Federal prosecutors allege that an OpenAI engineer hid an audio recorder behind a television in the conference room where the company’s leaders hold their most confidential talks, on business strategy, product plans and research. Abdulrahman Alfozan, 35, was arrested on 9 October. He has not been convicted of anything, and what follows is what the U.S. Department of Justice and the charging document allege, as reported by The Decoder on 14 October.

According to the Justice Department, the device was wrapped in toilet paper and pushed into a USB port at the back of the TV. Alfozan had been an OpenAI engineer since July 2025, and prosecutors say he had no access to that room. Investigators also found another recorder in his bag. OpenAI has fired him. A conviction would carry a prison term of up to five years and a fine of up to $250,000.

The part of the case that will draw the most attention concerns a chatbot. The charging document, as The Decoder describes it, says a chatbot account linked to Alfozan’s wife was used between 24 and 30 September to enter prompts. The questions touched on encrypting the SD card, tuning the Wi-Fi link, sharpening “audio and live listening,” and putting together a “continuous audio recorder.” Prosecutors say his own chatbot account has several identifiers in common with hers. They also say prompts naming the recovered gadget, down to its MAC address (the hardware ID a networked device carries), went in through the same account.

Which chatbot this was is not known. The Decoder says the filing does not name it, so there is no basis for saying it was ChatGPT, and the irony of OpenAI’s own product being the tool is not something the record supports. It is also an allegation about whose hands were on the keyboard. The account belongs to his wife, and prosecutors rely on shared identifiers and the content of the prompts to connect it to him. A court will test that link.

OpenAI’s statement is narrower than it may sound. The company said there are “no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data.” That is a company reporting what it has found so far, not a guarantee that nothing happened, and the statement as reported does not say how OpenAI looked or how far back.

FBI Director Kash Patel put the case in a larger frame, calling the protection of American “super intelligence innovations” “one of our highest counterintelligence priorities.” Neither the Justice Department’s account nor OpenAI’s statement, as relayed by The Decoder, says who if anyone besides Alfozan was involved, or what he meant to do with a recording. Those questions are open, and nothing in the record answers them.

The case matters for a reason that has little to do with the chatbot. A frontier lab’s security thinking is built around an outside attacker trying to pull out the crown jewels, which are the model weights and the code. Network monitoring and tight access to the training systems are aimed at that threat. The conduct alleged here is a different kind of problem: an insider, with a small recording device, in a room where the most valuable material is spoken aloud and never stored anywhere a weights-focused control would watch. That describes the shape of the exposure, not a motive, and the record offers none.

The access question is also unresolved. Prosecutors say Alfozan did not have access to the room, yet the device was allegedly found there. Neither the Justice Department’s account nor OpenAI’s statement, as The Decoder relays them, explains how it got in.

For other labs, the near-term consequence is mundane: sweep the executive conference rooms, and check what plugs into the television.

Reported by The Decoder on 14 October 2026, based on the U.S. Department of Justice press release and OpenAI’s statement.