OpenAI says an employee hid a secret recorder in its exec meeting room. Its own AI helped catch him

The device was held in place by a small piece of toilet paper.

It was plugged into a USB port on the back of a television in an executive conference room at OpenAI, the company behind ChatGPT. OpenAI's security team found it on 30 September. Two weeks later, the company told the world that the person who put it there was one of its own engineers.

OpenAI's announcement on X on 13 October has already passed 14 million views. The company says it used its own AI models to work out what the device was and where it came from. That led to one employee, who it says planted the device a few days earlier. The FBI arrested him. OpenAI fired him.

What did OpenAI find hidden in its office?

A small audio recorder, according to the FBI complaint reported by Fortune and CBS News. It was plugged into the back of a TV in an executive conference room in San Francisco.

OpenAI's technical surveillance countermeasures team found it during a sweep of the room. That is a team whose whole job is to look for bugs.

Fortune reports that the device had strong encryption and a way to send data out that was built to avoid being spotted. The FBI says the room was used to talk about confidential research, product plans and business deals, and sometimes with government officials.

The device was reportedly installed about two days before it was found.

How did OpenAI use its own AI to catch him?

OpenAI says it used its models "to understand the device and where it had come from." That is all the company has said about the method. It has not explained what the models actually did.

What it does say is the result. The trail ended at one employee, who it says planted the device a few days earlier.

Read nextPope Leo XIV on beauty, tech and why young people need to slow down

Then OpenAI called the FBI. In its post it thanked the agents and prosecutors "who moved quickly to make an arrest."

Who is the engineer accused of planting the bug?

Credible outlets including CBS News, Fortune and CNBC name him as Abdulrahman Alfozan, 35. Prosecutors say he worked at OpenAI as an engineer from July 2025. Reports describe him as a Saudi national. Fortune says he has lived in the US as a legal permanent resident since 2020.

He is charged with trying to use a device to intercept oral communications. If he is convicted, the maximum is 5 years in prison, according to Fortune.

These are allegations. His lawyer, William Pernik, says his client is presumed innocent. He also warned people not to politicise the case because of the man's nationality.

Photo: Sam Wheeler / Wikimedia Commons, CC BY-SA 3.0

What did he ask a chatbot, according to the complaint?

Here the story gets stranger. Fortune reports that the complaint says a chatbot account linked to his spouse was used from 24 September. The prompts allegedly asked how to build a recorder, how to encrypt what it records and how to listen in live. They also asked about building a second and a third device.

The complaint also says he used an internal OpenAI "productivity tool." CBS News reports that he asked it about building addresses, meeting rooms, room sizes, calendars and where leaders and outside partners would meet.

So one AI allegedly helped plan the bug. Another AI helped catch the person.

Photo: Disk Depot Ltd. / Wikimedia Commons, CC BY-SA 4.0

Day by day: How the bug was planted and caught

The dates below come from the complaint as reported by Fortune, CBS News and OpenAI's own post.

Did he get any of OpenAI's secrets?

OpenAI says there are no signs he "inappropriately" accessed or took model weights, source code or customer data. Model weights are the files that hold what an AI has learned. They are the crown jewels of a lab like this.

OpenAI also says its security program is built to stop everyone from careless insiders to criminals to foreign states. It calls this case an example of how AI can help against those threats.

Trending nowPetrol broke R30 a litre. We followed the money

The FBI's investigation is still going. Fortune reports it is unclear what he planned to do with any recordings, and that no leak has been found.

FBI Director Kash Patel called it a national security matter. He said protecting American "superintelligence" is one of the bureau's top counterintelligence priorities.

Photo: Crew crew / Wikimedia Commons, CC0

Why are people calling it a marketing stunt?

The reaction online was fast. Some joked that "John Connor got caught," a nod to the human hero of the Terminator films who fights the machines. Others picked on the words "The employee was terminated." Terminated is what the machines say in those same films.

Others were less amused. Many wrote that the post was "marketing in every situation," because OpenAI got to say its AI caught the culprit. OpenAI's own wording invites that. It ends its post by saying the case "is an example of how AI can help address" security threats.

The court papers, though, give the story a life of its own. An FBI complaint is not a press release.

OpenAI is no stranger to wild AI news. Our earlier stories cover the AI agents that formed a swarm and the later incidents that followed. This one is different. The danger here is a person with a USB device.

Where was he going three days after the arrest?

The last detail in the complaint, as Fortune reports it, is about travel.

The engineer was due to leave with his spouse and one other family member on a one-way flight to Saudi Arabia, via Istanbul. The flight was booked for three days after his arrest.

On 6 October he reportedly collected that bag and reported missing items, including a credit card, about $500 in cash, AirPods and sunglasses. He did not mention the second recorder.

Who was meant to hear the recordings? Nobody has said.