FBI Charges OpenAI Engineer With Hiding Recorder in Executive Room

US federal prosecutors have charged a 35-year-old OpenAI engineer with hiding a recording device inside the company's San Francisco headquarters — and OpenAI says its own AI models cracked the case. The defendant, Saudi national Abdulrahman Alfozan, faces up to five years in federal prison, a $250,000 fine and three years of supervised release if convicted.

According to the San Francisco Standard, which reviewed the criminal complaint, Alfozan had worked at OpenAI since July 2025. Prosecutors accuse him of trying to record oral conversations between OpenAI employees. The device, they say, sat in an executive conference room he wasn't authorized to enter or use.

That detail matters. Meetings in that room routinely covered confidential material, prosecutors said, including the company's product roadmap, research and government engagement. For a company at the center of US AI policy and racing competitors on model releases, an eavesdropping device in that specific room is a serious breach, not a quirky office incident.

What exactly did OpenAI find?

On September 30, OpenAI security found a DIY recording device plugged into a USB port on the back of a TV in the conference room. The concealment method was crude: prosecutors say Alfozan used toilet paper to hide it. OpenAI later determined the device went in sometime around September 28.

The company went public with the discovery in a statement on X on October 13. "We discovered a sophisticated recording device that was hidden inside one of our offices," OpenAI wrote. "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."

The twist — OpenAI investigating an alleged AI-assisted espionage attempt with its own AI — is the part the company chose to highlight. And it fits neatly into its argument that its models are useful analytical tools, not just chatbots.

The takeaway: OpenAI found the bug on September 30, traced it to an employee with its own models, and handed the case to the FBI.

How does the FBI say he built it?

The complaint lays out an uncomfortable irony: prosecutors allege Alfozan used a chatbot to plan the device. Investigators say he used a "productivity tool" to collect information about OpenAI's facilities. They also found prompts on a chatbot account tied to his spouse asking an AI model how to "build a continuous audio recorder" and "encrypt the SD card."

In other words, the same class of technology OpenAI sells allegedly helped design the recorder, and then helped OpenAI trace it. Prosecutors paint a picture of someone who asked an AI how to build, install and hide a remote recording device — and then did it.

The takeaway: the complaint alleges the recorder was planned with chatbot assistance, down to encryption of the storage card.

Was any data actually stolen?

OpenAI says it sees no evidence of the worst-case scenario. "At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data," the company wrote in its X statement. "The employee was terminated."

There's a second data point that complicates the picture. On October 5, according to prosecutors, Alfozan allegedly left a bag elsewhere in OpenAI's office. Security personnel searched it and found his wallet, his ID and a second recording device.

OpenAI did not immediately respond to Engadget's request for further comment.

The takeaway: OpenAI says no weights, source code or customer data left the building, but a second device turned up on October 5.

What happens next?

Alfozan made his first court appearance on Tuesday. The charges carry a maximum of five years in prison, a $250,000 fine and three years of supervised release — a relatively light ceiling for a case involving one of the most closely watched companies in AI.

The case lands at a moment when AI firms are tightening physical and digital security around model weights, research plans and government contracts. It also hands both sides of the AI-safety debate a fresh anecdote: a chatbot allegedly helped plan corporate espionage, and an AI model allegedly helped solve it. Expect the prosecution, and OpenAI's security practices, to face scrutiny as the case moves through federal court.