A company that trains AI to answer questions about nearly everything apparently needed its own AI to identify a piece of hardware found plugged into a TV in one of its conference rooms. That detail alone says something about where we are in 2026. According to Engadget, US federal prosecutors have arrested a 35-year-old Saudi national named Abdulrahman Alfozan, an OpenAI engineer who joined the company in July 2025, on charges of secretly recording conversations inside OpenAI’s San Francisco headquarters.
The alleged setup was not particularly high-tech, but it was deliberate. Alfozan is accused of building a DIY audio recording device and hiding it inside an executive conference room he had no authorization to access or use. The device was found on September 30, plugged into a USB port on the back of a TV and partially covered with toilet paper. A second recording device was discovered on October 5, tucked inside a bag Alfozan allegedly left elsewhere in the office, along with his wallet and ID. That second find was, to put it plainly, a significant operational mistake.
The detail that should make anyone think twice is how prosecutors say Alfozan planned this. A chatbot account linked to his spouse allegedly contained prompts asking an AI model how to “build a continuous audio recorder” and “encrypt the SD card.” He also reportedly used a productivity tool to research OpenAI’s office layout. So the same category of AI product OpenAI sells to the public may have been used to plan corporate espionage against OpenAI itself. The irony is hard to miss, and the privacy implications run in multiple directions.
OpenAI said in a post on X that it used its own AI models to “understand the device,” which it described as sophisticated. The company also stated it has “no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data.” Alfozan was terminated. OpenAI did not respond to a request for additional comment from Engadget.
This matters beyond the drama of a spy story. The meetings held in that conference room reportedly covered OpenAI’s product roadmap, ongoing research, and government engagement. That’s a significant target profile. OpenAI is not a neutral tech company at this point. It has contracts with the US government, active relationships with defense and intelligence-adjacent agencies, and research that foreign governments have clear incentives to monitor.
Alfozan made his first court appearance on Tuesday. If convicted, he faces up to five years in federal prison, a $250,000 fine, and three years of supervised release. But the bigger question this case raises has nothing to do with sentencing. It’s about whether organizations handling genuinely sensitive research are doing enough physical security, not just cybersecurity, to protect what’s inside their walls. A USB device hidden behind a TV suggests the answer, at least in this case, was no.