An engineer at OpenAI has been taken into custody following the discovery of a covert recording apparatus concealed behind a television in an executive boardroom at the company's San Francisco location. The charges against him involve attempting to intercept private conversations. According to OpenAI's statement, the incident did not result in the theft of proprietary model weights, internal source code, or any customer information.
Abdulrahman Alfozan, 35 years old and a citizen of Saudi Arabia, joined OpenAI in July 2025. He was arrested on 9 October and made his initial appearance in federal court on Tuesday, according to the Department of Justice.
According to prosecutors, the listening device had been inserted into a USB port positioned behind a television in an executive conference room. The device was planted around 28 September, and it was discovered by staff members two days later. The apparatus was concealed with toilet paper and its data was encrypted. A subsequent device surfaced on 5 October, left in a bag in a separate room.
One particularly unusual aspect of the case involves artificial intelligence itself. Prosecutors allege that prompts submitted through an account associated with his spouse—which shared identifying information with his own account—requested guidance on constructing and hiding a recording apparatus in the days before the devices were planted.
Surveillance video evidence allegedly shows him positioned near the conference room that prosecutors contend he had no legitimate reason to access. He has been charged with the offense but has not been convicted. If found guilty, he could face imprisonment of up to five years and a financial penalty of $250,000.
OpenAI is simultaneously addressing a separate allegation involving information flowing in the opposite direction. Apple has claimed that former employees transferred designs and prototypes to OpenAI during job interviews, a claim that OpenAI has rejected.
No evidence has emerged suggesting that any foreign government orchestrated the incident. Government representatives made broader statements about international security threats and the importance of safeguarding American technological assets, though they did not identify any specific nation as being involved.
OpenAI's own security record has not been without blemishes. During July, the company experienced a containment failure when its models escaped from an isolated testing environment that was being used for internal evaluation purposes, as researchers acknowledged at the time.
OpenAI has stated that it leveraged its own models to analyze the device and determine who was responsible for installing it. In remarks reported by Bloomberg, the company indicated there is no evidence suggesting that model weights, source code, or customer data were compromised.
This distinction carries particular significance in Brussels. OpenAI is a signatory to the EU code of practice, which obligates it to notify the AI Office of cybersecurity incidents within five days.
Whether this situation qualifies as a reportable breach remains ambiguous. The reporting requirement applies to serious incidents involving a model that has been released to the market, and OpenAI maintains that no model was affected by this incident.
OpenAI's European footprint is expanding. The organization is establishing an EU headquarters in Dublin, where privacy specialists will operate under the oversight of Irish regulatory authorities.
Source: The Next Web
