Somewhere around September 28, according to federal prosecutors, a small recording device was wrapped in toilet paper and pushed into a USB port on the back of a television in one of OpenAI’s executive conference rooms in San Francisco. Two days later, the company found it. Ten days after that, the FBI arrested the engineer they say put it there.
The man is Abdulrahman Alfozan, a 35-year-old Saudi national who worked at OpenAI. He made his first appearance in federal court on October 13 and remains in custody. And the detail that has made the case spread so fast online is not the device itself. It is how prosecutors say he built it: by asking a chatbot.
The short version
- Who: Abdulrahman Alfozan, 35, an OpenAI engineer, now fired
- What: accused of planting a covert audio recorder behind a TV in an executive conference room
- The AI twist: prosecutors say chatbot prompts covered building, hiding, encrypting and remotely accessing the recorder
- The charge: one count of intercepting or trying to intercept oral communications, up to 5 years in prison and a $250,000 fine
- What OpenAI says: no sign that model weights, source code or customer data were accessed or taken
How the device was found
OpenAI disclosed the incident itself in a post on X. “We discovered a sophisticated recording device that was hidden inside one of our offices,” the company 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 Justice Department’s account fills in the rest. According to prosecutors, the device was installed around September 28 in a USB port on the back of a television in an executive conference room, wrapped in toilet paper. It was discovered on September 30. Security camera footage showed Alfozan near the room on several occasions, and on October 5, prosecutors say, he came back to check on the device. A photo in the complaint reportedly shows him entering the room and looking behind the TV.
That same day, OpenAI security found an unattended bag elsewhere in the building. Inside, according to prosecutors, were his wallet, his ID and a second recording device. He was arrested on Friday, October 9, and OpenAI fired him the same day.
The chatbot prompts
This is the part of the case that sounds like satire but is laid out in the charging documents. According to prosecutors, the prompts covered “building, customizing, installing, audio recording, testing, obfuscating, and remote accessing a device.” The complaint quotes requests to “build a continuous audio recorder,” “customize Wi-Fi … connectivity,” “improve audio and live listening” and “encrypt the SD card.”
Reports on the complaint say the prompts came from a chatbot account linked to his spouse that shared identifiers with his own account, and that some of them referred to the specific device later found, including details that matched it. The complaint does not name which chatbot was used, and prosecutors have not accused the spouse of writing the prompts.
Separately, the FBI says he used an internal “productivity tool” to look up OpenAI building addresses, meeting rooms and executive information to decide where to place the equipment. FBI Director Kash Patel said Alfozan “allegedly concealed a covert recording device inside his employer’s executive conference room to secretly capture sensitive communications.”
Why this matters for every chatbot user
Prompts are records. In this case they reportedly became some of the clearest evidence in a federal complaint. Anything you type into an AI assistant can be stored, reviewed and, with the right legal process, handed to investigators. We broke down what OpenAI keeps and who can read it earlier this year.
What he is charged with and what he faces
Alfozan is charged by criminal complaint with one count of unlawful interception, or attempted interception, of oral communications under 18 U.S.C. § 2511(1)(b), part of the federal Wiretap Act.
| Item | Details |
|---|---|
| Charge | Intercepting or attempting to intercept oral communications, 18 U.S.C. § 2511(1)(b) |
| Maximum prison term | 5 years |
| Maximum fine | $250,000 |
| Supervised release | Up to 3 years |
| Not charged (so far) | Theft of trade secrets, economic espionage |
| Next court date | October 19, detention hearing in the Northern District of California |
Maximum penalties are rarely what a defendant actually receives, and actual sentences depend on the federal guidelines and the judge. More charges are possible if investigators find evidence of other recordings or of anyone receiving the data, but none have been filed.
His lawyer, William Pernik, has told reporters that his client is presumed innocent, cautioned against “jumping to conclusions” so early, and warned against politicizing the case because of his client’s nationality.
What is still unknown
The biggest open question is why. Prosecutors have not said what Alfozan planned to do with the recordings, and they have not alleged that he was working for a foreign government, a competitor or anyone else. No buyer or recipient is named in the complaint. The second device has only been described as a recording device, and the contents of the first were encrypted.
OpenAI has tried to limit the damage. “At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data,” the company said. “The employee was terminated.” It added that its security program “is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats.”
That last line is worth reading carefully. OpenAI says its models helped trace the device, but it has not explained how, and the evidence described in the court filings, such as camera footage, internal tool searches and chatbot logs, is the kind of trail investigators have always used.
Not the first AI insider case
AI labs have been warning for years that their most valuable assets, trained model weights and the research behind them, are targets for insiders as well as outside hackers. A RAND study on securing model weights listed dozens of distinct attack routes, including insider access and human intelligence operations. Most of the recent cases, though, have looked very different from a bug behind a TV.
| Case | What was alleged | Where it stands |
|---|---|---|
| OpenAI / Alfozan (2026) | Covert audio recorder in an executive conference room | Charged by complaint; detention hearing October 19 |
| Google / Linwei Ding | Copying AI supercomputing and data center files | Convicted in January 2026; espionage counts later thrown out; sentenced to about a year |
| xAI vs. OpenAI | Trade secrets taken through hiring of former staff | Dismissed with prejudice in June 2026 |
The Alfozan case also lands during an uncomfortable month for OpenAI’s internal security. Days earlier, the company fired three safety researchers over what it called the mishandling of sensitive information, a claim the researchers dispute. The two cases are unrelated, but together they show how tightly OpenAI is now guarding what happens in its rooms.
What happens next
The October 19 hearing will decide whether Alfozan stays in custody until trial. Prosecutors will have to move from a complaint to an indictment, and that filing could add charges or more detail about motive. Until then, the case rests on the allegations: a device behind a TV, a bag left in the wrong room and a set of chatbot prompts that, if prosecutors are right, read like a step-by-step plan.
The FBI framed the case in national security terms, with Patel calling the protection of American AI work “one of our highest counterintelligence priorities.” Whether the evidence ends up supporting that framing is the part nobody can answer yet.
Sources and further reading
- UNILAD Tech: Prison time OpenAI employee could face after being caught planting recording device in exec offices
- U.S. Attorney’s Office, Northern District of California: Saudi national arrested for installing surreptitious audio recording device at OpenAI
- CNBC: OpenAI employee arrested over recording device
- TechCrunch: FBI arrests OpenAI employee accused of planting hidden recording device
- CBS News: OpenAI employee arrested
- Semafor: Saudi national arrested over alleged planting of recording device at OpenAI
About this article: GeekBlog covers U.S. technology news, AI, phones, smartwatches and gaming. Every story is written and checked under our Editorial Policy. Spotted a mistake or have a story tip? Contact our editors.
