A 35-year-old citizen of Saudi Arabia who was an employee at OpenAI was accused by federal prosecutors of installing a hidden audio recording device at the artificial intelligence company’s headquarters in San Francisco.
Abdulrahman Alfozan is accused of using an electronic, mechanical, or other device to intercept oral communications, according to the U.S. Attorney’s Office for the Northern District of California.
The device was allegedly discovered on Sept. 30 by a security team at OpenAI. The audio recording device was allegedly found during a sweep of office spaces and was inserted into the USB port on the back of a TV in an executive meeting room, according to OpenAI.
The company was one of the first artificial intelligence-backed businesses to roll out chat and search products using AI in late 2022. Its most well-known product is ChatGPT, which is a large language model that uses AI to chat with users and generate text, images, and other content.
“We discovered a sophisticated recording device that was hidden inside one of our offices,” an OpenAI spokesperson said in a statement Tuesday. “We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and we’re grateful to the agents and prosecutors who moved quickly to make an arrest.”
The device was designed to evade detection through advanced encryption and was capable of recording audio and transferring data, according to OpenAI.
The federal criminal complaint alleges the device was wrapped in toilet paper to try to conceal it.
An internal investigation allegedly suggested the device had been placed there by Alfozan, who had been employed as an engineer on the product team at the company since July 2025, according to U.S. Attorney Craig Missakian’s office.
The device was allegedly installed on Sept. 28 before being found two days later. The company said it found evidence that Alfozan planned to install additional devices.
On Oct. 5, another device was allegedly found in a bag Alfozan had left in a printer room, according to the criminal complaint. The complaint alleges that when Alfozan retrieved the bag the next day from the building’s lost and found, he asked about items missing from the bag but did not ask about the audio recording device.
The complaint also alleges that Alfozan made several inquiries on an unnamed “productivity agent” about OpenAI’s building, such as its layout, address, and meeting schedule. His wife’s account was allegedly used for other queries about how to build an audio recording device.
His employment was maintained until the FBI investigation was concluded, with the company gradually reducing his access to space and equipment before suspending his access altogether and placing him on leave. He was terminated on Friday, the same day he was arrested by the FBI, according to OpenAI.
Alfozan made an initial court appearance Monday in San Francisco and will next appear before U.S. Magistrate Judge Alex Tse next Monday.
In a statement, Missakian said the issue was being treated as a threat to national security.
“The Bay Area and Silicon Valley lead the world in developing tomorrow’s technology, and our office will continue to stand at the forefront of protecting that American innovation,” Missakian said in a statement. “For us it is a matter of national security. This arrest highlights our relentless efforts in partnership with the FBI to identify and prosecute those who seek to steal our technology, as alleged in the complaint.”
The federal criminal complaint and the press release from the U.S. Attorney’s Office for the Northern District of California use the terms “Super Intelligence” and “SI,” which were recently suggested as replacement terms by President Donald Trump for artificial intelligence and AI.
An OpenAI spokesperson did not address questions on whether the company uses the term to refer to its technology. A free version of ChatGPT, the company’s flagship product, returned an answer that the term Super Intelligence is not accurate.