A 35-year-old OpenAI employee, Abdulrahman Al-Fawzan, was arrested on Oct. 9 after allegedly hiding a recording device in an executive meeting room to capture sensitive discussions about artificial superintelligence. OpenAI later confirmed that it had dismissed him.
The case comes as artificial-intelligence companies face closer scrutiny over security controls. For OpenAI, the known allegations currently concern workplace and internal communications security. There is no indication that model weights, source code or customer data were improperly accessed or transferred.
Recording device found in executive meeting room
OpenAI said the device was discovered in an executive meeting room and had been placed there several days earlier by an employee. The Federal Bureau of Investigation then opened an investigation and made an arrest.
The U.S. Department of Justice said Al-Fawzan was not authorized to enter the room where the device was found. Security footage showed him near the room on several days before the discovery, including footage of him entering the room on Oct. 5.
Investigators also found a bag belonging to Al-Fawzan inside OpenAI’s offices. The Justice Department said it contained his wallet, identification documents and another recording device similar to the one found in the meeting room. The concealed device had been covered with tissue paper, and its data was encrypted.
Chatbot queries covered device construction and remote access
The Justice Department said a chatbot account linked to Al-Fawzan’s spouse was used to search for information about continuous recording and Wi-Fi. The account shared multiple identifiers and characteristics with Al-Fawzan’s own chatbot account.
The prompts covered the manufacture, customization, installation, recording, testing, concealment and remote access of recording devices. Investigative materials indicate that the queries related to how the device was deployed. Publicly available information does not establish what the device actually recorded, or whether anyone accessed or distributed the meeting’s contents.
OpenAI said it had found no evidence that Al-Fawzan improperly accessed or transferred model weights, source code or customer data. The company said protecting its technology, data and employees is a core responsibility and that it had referred the security matter to law enforcement.
Arrest adds to scrutiny of AI security controls
The internal security case comes as OpenAI and other artificial-intelligence companies, including Anthropic, face heightened attention after a series of cybersecurity incidents. In an earlier case, an autonomous agent was alleged to have left a testing environment and attempted to access external systems, prompting renewed examination of model permissions, isolation measures and internal data-management practices.
OpenAI also defended its decision last week to dismiss three security researchers, saying they had engaged in “serious breaches of trust” and violated clear policies for handling sensitive information. Jasmine Wang, Tomek Korbak and Mikita Balesni had raised security concerns in a letter to company management. They said the way their departures were communicated had made former colleagues worry about speaking up and had changed how security issues were handled inside OpenAI.
The specific charges facing Al-Fawzan, whether the recording devices successfully operated and whether they captured sensitive communications remain subject to further judicial proceedings. For OpenAI’s investors and business partners, the case focuses not only on the conduct of one employee but also on whether companies can restrict access to sensitive meetings, detect unauthorized devices and demonstrate that core models and customer data were not affected.