Espionage

A Bug and Stolen Files: OpenAI Is Fighting Espionage on Two Fronts

co-produced by AI14. October 2026, 18:56

First a bug in the executive conference room, then a former employee who allegedly took more than 100 confidential files to Apple: OpenAI has a serious problem protecting its secrets. Within a few days, two espionage allegations involving the ChatGPT maker have come to light. In the first case, the F.B.I. arrested an OpenAI engineer accused of hiding a recording device in the leadership team’s conference room. In the second, OpenAI accuses a former employee of taking documents about an unreleased device and presenting them at Apple. In both cases, the accused are presumed innocent.

Case 1: The Bug Behind the TV

Last Friday, the F.B.I. arrested Abdulrahman Alfozan, a 35-year-old Saudi national who had worked as an engineer on OpenAI’s product team since July 2025. Before that, he spent about five years at Meta. According to the U.S. Attorney’s Office for the Northern District of California, he plugged a hidden recording device into a USB port on the back of a TV in the executive conference room in San Francisco in late September and disguised it with toilet paper. According to the F.B.I., highly confidential topics are discussed in that room, including research, product plans and relationships with governments and business partners.

The charges paint a detailed picture:

The preparation: Mr. Alfozan allegedly used an internal tool to search for meeting rooms, calendars and the locations of leadership meetings.
Help from a chatbot: Using a chatbot account registered to his wife, he allegedly asked how to build a continuous audio recorder, hide it, listen in remotely via Wi-Fi and encrypt the memory card.
The cameras: Surveillance footage allegedly shows him near the room, to which he had no access, on several days.
The second device: In a bag Mr. Alfozan later left in another room, security staff found a second, similar recording device.

The bug was discovered during a routine sweep just two days after it was allegedly installed. OpenAI says it also used its own A.I. models to analyze it. According to OpenAI, there is no evidence that Mr. Alfozan accessed model weights, source code or customer data. Prosecutors have not named a motive, nor have they alleged that he acted on behalf of a government or a competitor. If convicted, he faces up to five years in prison. Kash Patel, the F.B.I. director, called protecting “superintelligence technology” one of the bureau’s top counterintelligence priorities.

Case 2: To Apple With More Than 100 Files

The second case is part of the legal battle between OpenAI and Apple. In a new court filing, OpenAI accuses a former employee who now works at Apple of sending more than 100 files about an unreleased product to a personal account, Reuters reported. The employee later allegedly gave a presentation about the product to Apple’s senior leadership. The report does not name the employee or the product, which is said to be a device. Apple has not yet commented.

The allegation is notable because Apple is actually the plaintiff. In July, the iPhone maker sued OpenAI and its hardware company io Products, which OpenAI built with the former Apple design chief Jony Ive, for theft of trade secrets. Apple accuses OpenAI of deliberately recruiting former Apple employees to obtain confidential information about product design, manufacturing and the supply chain. At the center of the case are OpenAI’s hardware chief, Tang Tan, a former Apple vice president, and the engineer Chang Liu, who allegedly downloaded confidential files. OpenAI is also accused of asking job candidates to bring prototypes and design documents to interviews.

OpenAI denies the allegations. It says Apple merely wants to stop employees from moving to competitors and is treating ordinary job interviews as trade secret theft. Apple has not identified any specific stolen trade secrets, OpenAI argues, and it has moved to dismiss the lawsuit. On Tuesday, Judge Edward Davila of the U.S. District Court in San Jose ordered OpenAI to quickly hand over documents about job interviews with former Apple employees. Today, he is hearing Apple’s request for a preliminary injunction and OpenAI’s motion to dismiss.

OpenAI’s First Device Is at Stake

Behind the dispute lies a market worth billions. OpenAI is developing its own hardware, which Sam Altman says will create a new class of devices beyond the smartphone. According to Bloomberg, the first product will be a doughnut-shaped smart speaker priced at more than $300. That makes OpenAI a direct competitor to Apple, even though the two companies are partners: ChatGPT has been integrated into Apple’s software for two years. In May, however, OpenAI was already considering legal action against Apple because the ChatGPT integration was not going as hoped.

Not an Isolated Case in the A.I. Industry

In early October, OpenAI parted ways with three members of its safety team who allegedly leaked confidential information (Trending Topics reported). Suspicions of espionage are nothing new in the industry: In 2024, a former Google engineer was charged in the United States with stealing trade secrets about A.I. chips for Chinese companies. And in 2025, Elon Musk’s xAI sued a former employee who allegedly took confidential documents before moving to OpenAI. The more valuable the technology, the fiercer the fight over people and secrets.