OpenAI found a hidden recording device in an executive conference room at its San Francisco office and called in the FBI. The employee, named by federal prosecutors as Saudi national Abdulrahman Alfozan, was fired and arrested on October 9, about a week after the discovery. He started in July 2025 and worked on the product team, an OpenAI spokesperson confirmed to TechCrunch.
A camera behind the television cracked the case
According to prosecutors with the US Attorney’s Office for the Northern District of California, the device was inserted into a USB port on the back of a TV. That is exactly where OpenAI had placed a camera, and it allegedly caught Alfozan looking behind the set. The charge is using an electronic device to intercept oral communications, meaning recording people’s voices without permission. A conviction carries up to five years.
OpenAI says its own internal models helped it understand the device. The company calls it “sophisticated” and lists advanced encryption, audio monitoring capabilities and a data transfer mechanism designed to evade monitoring. How far that actually sets it apart from other recording hardware is unclear.
The second half of the company’s statement is the part customers care about: “At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data.”
The espionage framing comes from the FBI, not from OpenAI
What the recordings were for is still unknown. FBI director Kash Patel says the scheme was to “capture sensitive communications about the company’s super intelligence technology”, and he files the case under counterintelligence: protecting American SI innovation, he said, is one of the bureau’s highest priorities. OpenAI itself is markedly more careful and names no motive at all.
The gap between those two readings matters. A fired employee with a recording device is an incident any company can have. State-directed industrial espionage inside an AI lab’s meeting room would be a different story, and nobody has shown that yet. Neither Alfozan nor his attorney could be reached by TechCrunch.
Here’s what sticks regardless of motive. Security debate inside AI labs has run for months on model weights, API keys and distillation. The leak that actually turned up here was a microphone in a meeting room.