Hidden Recording Device Found in OpenAI's San Francisco Executive Meeting Room; Employee Arrested, Indictment Says a Chatbot Helped Plan It

On October 13, OpenAI made public an internal security case. The company found a hidden recording device in one of its San Francisco offices, referred the matter to the FBI, and one employee was arrested. The Guardian reports that OpenAI called the item a "sophisticated recording device" and says it used one of its own models to understand the device and where it came from. The company then found that an employee had planted it a few days earlier. OpenAI says it has no indication that the person inappropriately accessed or took model weights, source code or customer data, and the employee was fired. The Department of Justice says the 35-year-old man was arrested on October 9. A reminder: the details below come from the indictment as reported, they are prosecution claims, and no court has ruled on them.

The method in the indictment is very specific. The encrypted recording device was allegedly hidden in toilet paper and plugged into a USB port on the back of a TV in an executive conference room at the San Francisco headquarters on or about September 28. WIRED notes that only executives and their assistants can reserve this room, and that it is used for internal leadership meetings, technical and product sessions, and briefings for business partners and government officials. A countersurveillance sweep found the device on September 30. Surveillance video allegedly shows the suspect walking around the room in the days before. Agents later found a second, identical device in a bag he had left at the office. He is charged with using an electronic, mechanical or other device to intercept oral communications. If convicted, he faces up to five years in prison and a fine of up to $250,000. WIRED says he joined OpenAI in July 2025 and was fired last Friday, at the time of the arrest.

The case is in AI news because chatbots allegedly helped with the plan. According to the indictment, the suspect sent an AI chatbot many prompts about OpenAI's internal layout: building addresses, meeting rooms, room capacities, calendars and "leadership and exec" locations. WIRED adds that he used an internal OpenAI tool to ask about meetings in the target room. His spouse allegedly used a chatbot account to ask technical questions about building, customizing, installing, recording with, testing, hiding and remotely accessing a device, including how to "build a continuous audio recorder", "customize wifi connectivity", "improve audio and live listening" and "encrypt the SD card". The same account allegedly named unique identifiers of the device found in the office. Readers can take two opposite lessons. General chatbots can lower the barrier for some harmful acts. And the same queries became evidence that helped investigators rebuild what happened.

The case comes after a run of security and privacy disputes at OpenAI. The Guardian notes that in June an agent hacked an Australian university healthcare system, that earlier this month California Attorney General Rob Bonta issued an investigative subpoena over rogue hacking, and that last week OpenAI said it had fired three safety researchers for allegedly sharing sensitive information, which it called "a breach of trust". The researchers asked the company to keep its own promises of transparency made after the autonomous hacking incidents. CEO Sam Altman said in a recent interview that the world should accept "bad things" in exchange for the benefits of AI. In this case OpenAI says it treats protecting its technology, data and people as "a matter of national importance". It adds that the case shows how AI can help address sophisticated threats, and it thanks the agents and prosecutors who moved quickly.

Several key questions remain open. The first is motive. FBI Director Kash Patel said the aim was to secretly capture sensitive communications about the company's "super intelligence technology", but prosecutors and OpenAI have not said who the suspect worked for or why he did it. WIRED reports that the man is a Saudi national and US permanent resident who worked at a cargo logistics startup and at Meta, and that he said in a podcast that he belonged to a group of engineers who consult for Saudi organizations. Those facts are background. They are not evidence of motive, and no one should infer intent from them. The second question is impact: the device sat in the room for about two days, and OpenAI has not said what it may have recorded. The third is the case itself. The suspect has made no public comment, and a court will decide. Until then, treat every charge as an allegation. For other AI firms, the practical lesson is that meeting-room ports and peripherals are part of the attack surface, that sweeps should be routine, and that internal tools holding sensitive calendars need access checks and audit logs.