How the device was found

OpenAI's internal surveillance team found a USB recording device on 30 September in an executive conference room at its San Francisco office. According to the FBI, the device was wrapped in toilet paper and plugged in behind a television in that room, and it was likely installed on 28 September. Its stored data was encrypted and could not be accessed. OpenAI said it used its own models to analyze the device and trace where it came from, which led to one employee who had placed it days earlier. On 5 October the company installed a camera in the room, and the footage allegedly shows the employee trying to retrieve the device that same day. The FBI also reported finding a similar device in the employee's personal bag.

The arrest and the charge

Abdulrahman Alfozan, a Saudi national and US legal permanent resident since 2020, was arrested on 9 October, the same day OpenAI ended his employment. He had worked at the company as an engineer since July 2025. The FBI announced the arrest in a press release on Tuesday, 13 October, when the criminal complaint was also unsealed. Alfozan made his first court appearance that Tuesday on a charge of using or endeavouring to use an electronic, mechanical or other device to intercept oral communications. If convicted, he faces up to five years in prison, a $250,000 fine and three years of supervised release. He is due back in federal court on 19 October, and it was not immediately clear whether he had a lawyer.

FBI director Kash Patel celebrated the arrest and cited a Trump memo calling for coordinated efforts to maintain America's AI advantage. He said the FBI treats this work as one of its highest counterintelligence priorities.

It's essential to safeguarding our nation's economic strength and national security, and the FBI views that mission as one of our highest counterintelligence priorities.

What the court filings describe

According to the complaint, between 17 August and 16 September the employee asked a productivity AI assistant several questions about meeting rooms and their availability. Some of the questions asked how outside partners would usually meet with OpenAI and which meetings were scheduled in that room. In September, the complaint says, he used a chatbot linked to his spouse to ask how to build and install a recording device and how to reach it remotely. CCTV recorded him near the conference room on eight separate occasions in September, although he had no authority to access or take part in meetings there. The room was regularly used for discussions of highly confidential technical, commercial and strategic information, along with government engagements and commercial relationships. A court filing also shows that a one-way ticket to Saudi Arabia had been booked for 12 October.

OpenAI's response

OpenAI said it has no evidence that this employee improperly accessed or extracted model weights, source code or customer data. It described protecting its technology, data and people as a matter of extreme importance, including at the national level. The company said its security program is designed to defend against a broad range of threats, from insiders to criminal hackers and state actors, and that the case shows how AI can help address such threats. Rather than removing him at once, OpenAI gradually cut his access to systems while it worked with the FBI to analyze the device, and it timed the firing with the arrest. Ars Technica reported that the company waited until enough evidence had been gathered before acting.

Open questions

Neither prosecutors nor OpenAI have given a motive for the alleged plot. Patel said Alfozan aimed to capture information about OpenAI's super intelligence technology. According to a LinkedIn profile in his name, he studied at MIT and UC Berkeley and previously worked at Meta and Microsoft, and Al Jazeera could not independently verify those details. The BBC reported that it is not yet clear what he planned to do with the recordings. The FBI investigation is ongoing, and the next court date is 19 October.