Someone Hid a Listening Device in OpenAI's Office
OpenAI said on October 13, 2026 that it found a hidden recording device in one of its offices and had fired the worker accused of putting it there. Federal prosecutors arrested and charged the former worker that same day. Engadget Reuters
The accused is Abdulrahman Alfozan, 35, an engineer who started at OpenAI in July 2025. The FBI arrested him over allegedly hiding a listening device in OpenAI's San Francisco headquarters. OpenAI shared the news on its OpenAINewsroom X account on October 13. OpenAINewsroom
Prosecutors say Alfozan left a homemade recorder in a meeting room for senior leaders that he was not allowed to enter or use. OpenAI said it was placed in and around September 28. Workers found it on September 30.
The placement was low profile. The device was plugged into the back of the room's TV, much like a phone charger left plugged in, and covered with toilet paper. OpenAI described the electronics inside as sophisticated.
The case did not end with that find. On October 5, security staff found Alfozan's wallet, ID and a second recorder in a bag he allegedly left in the office. That second find tightened the link between him and the devices.
OpenAI said it used its own AI tools to study the recording device and to help figure out who placed it. Business Insider
For customers and staff, OpenAI said it saw no signs that he had reached or taken model weights, the private files that shape how the AI acts, computer code, or customer data. This was secret audio recording in a meeting room, not a confirmed break-in to its computer systems or training computers.
If convicted, Alfozan faces up to five years in federal prison, a $250,000 fine and three years of supervised release.
The broader context here is one I have seen before in security work. The locked computer files are often not the easiest target. It is the room where people talk openly about plans, staff changes and safety issues. A small recorder behind a TV can keep running for days without setting off computer security alarms.
In my view, two points matter more than the toilet paper detail. Building access was not the same as permission to enter that room, yet the device still stayed there for about two days. And OpenAI used its own AI, alongside normal security work, to study the hardware and trace who left it.
Looking ahead, the fix is basic but important. Limit who can enter sensitive meeting rooms, check shared TVs and meeting gear by eye, and track what is plugged in. For a field focused on protecting code and data, microphones still count.
