A former OpenAI engineer has been accused of planting a listening device in a company meeting room, and on October 19, 2026, a US federal court ordered him held in custody while the case proceeds. Ars Technica reported that day, citing court filings, that federal prosecutors argued in an October 18 filing that the defendant, Abdulrahman Alfozan, has every incentive and the means to flee; Magistrate Judge Alex G. Tse agreed at Monday's hearing and denied release. One point upfront: these are allegations, not a conviction.
The evidence chain prosecutors laid out
According to the filings, OpenAI discovered a recording device hidden in a meeting room about two weeks before Alfozan's flight booking; the device had been in place for at least several days, including during a highly sensitive meeting with US government officials held in that room. The complaint says Alfozan used OpenAI's own productivity agent to research which meeting rooms would be best to exploit, and used his wife's ChatGPT account to research how to modify a covert recording device so it would record automatically when it detected voices and could be controlled remotely. After searching his home, the FBI said it found a recording device similar to the office one; his iPhone allegedly held photos of another covert recording device, photos of input and output panels on electronics around the OpenAI building including another sensitive meeting room, and a folder named Saudi Gov Docs that he accessed on October 4 — the folder's contents have not been made public. Prosecutors also say WhatsApp logs show a member of the Saudi royal family in contact with him from 2025 onward, including a message telling him Saudi Arabia needed him more than the US, to which he replied 100.
Why the court sees a flight risk
Prosecutors say Alfozan had booked a one-way ticket to Riyadh paid for by the Saudi government, and that the FBI arrested him days before departure and seized his Saudi passport. After his arrest, his wife and son rebooked onto an earlier flight and left the United States; prosecutors say their one-way business-class ticket to Doha was booked and paid for by a member of the Saudi military. Alfozan is a Saudi national and US permanent resident who moved to the US in 2009, earned technical degrees at MIT and UC Berkeley, worked in Silicon Valley for over a decade, and joined OpenAI around July 2025. He is currently charged with illegally recording OpenAI's meetings, which carries up to five years in prison; prosecutors say more serious charges may follow as evidence grows. The Ars Technica report carried no direct response from OpenAI.
Frontier labs now have an insider problem to design for
The sharp edge of this case is not the bugging technology but the tooling: the reconnaissance was allegedly done with the company's own productivity agent and ChatGPT. AI tools that can reach internal information and also answer how-to questions are becoming a new channel for insider threats. For frontier labs, a meeting-room sweep now has to cover agent logs as well as wall corners: who asked which rooms host what meetings, and who researched modifying recording gear, are warning signals in themselves. It is the mirror image of the AI agent hacker who breached more than 100 websites covered on this site: agents can probe tirelessly for outside attackers, and can just as quietly speed up reconnaissance for someone on the inside. Where the case goes next depends on whether prosecutors can substantiate the motive allegation of acting for the Saudi government — prosecutors themselves concede that even if they cannot, the criminal character of the alleged bugging does not change.
