On 13 October 2026 the US Department of Justice announced the arrest of an OpenAI engineer. According to the criminal complaint, he planted a covert audio recording device in the company's executive conference room in San Francisco. We checked what a background check from public sources would have found, before he was hired and while he worked there.
According to the complaint, company staff found the device at the end of September 2026, plugged in behind a television in the room and wrapped in toilet paper. He had worked at the company as an engineer since about July 2025, and had no authority to enter the room. The Department of Justice does not allege that a foreign government was involved. OpenAI said it has seen no sign that model weights, source code or customer data were taken, and that he was fired.
As in our earlier case study on flight FZ1073, we fixed the method before searching. We checked two dates: the day before he was hired, and the day before the device was installed. We read the findings through two separate lenses: the US government's conduct-based insider-risk indicators, and ProfCheck's method. We also ran 23 Google searches limited to pages dated before the hiring. The search covered English and Arabic.
What we did not find
No criminal record, no sanctions, no hostility to an employer and no support for violence. We read 262 of the 284 posts on his X account, and none of them fits the negative behavioural indicators in our method.
Two red flags, public before he was hired
The first: employment dates contradicted by a record. On his LinkedIn profile, his work at Meta ends in 2022. But in January 2025 an internal Meta code change was published in Meta's open-source repositories, from a Meta email address, under his username. The record was public about five months before he joined OpenAI. Google does not surface it. It is found only by searching his code history on GitHub.
The second: an outside role alongside the job. In February 2025 he joined a Riyadh venture capital fund as a venture partner, and the role continued while he worked at OpenAI. The role was announced publicly, so it was not hidden from the public. Whether he disclosed it to his employer is not known.
Under our method, a contradiction in employment dates and the non-disclosure of a conflict of interest, if that occurred, are indicators of unreliability. They do not prove wrongdoing. They are red flags that call for a clarification with the candidate before a decision: an explanation of the dates, a written list of outside roles, and verification with the former employers. Whether that clarification would have changed the decision cannot be known.
Where the warning signs about the act itself were
According to the complaint, every warning sign about the act was inside the building:
• Cameras recorded him near the room on ten different dates in September.
• He searched the company's own systems for where and when the leadership meets.
• An account linked to him asked a chatbot how to build a covert, encrypted recorder.
And in my view this part matters even more. A background check from public sources helps decide whom to hire. It cannot see what an employee does once inside. That is what an internal insider-risk program sees: access control, cameras and the logs of internal tools.
What this means for employers
In practice, this means, among other things:
• Verify employment dates against records, not only against the CV and the LinkedIn profile.
• For technical candidates, check the public history of their code as well. It carries dates and employer domains.
• Ask in writing, at hiring and every year, for a list of outside roles, investments and advisory work. A role announced in public is not necessarily a role disclosed to the employer.
• Treat a red flag as a question that needs an answer, not as a verdict.
• In an organization that holds sensitive information, do not rely on the check at the door alone. A background check and an internal insider-risk program do different jobs, and both are needed.
It is important to note: he has been charged in a criminal complaint, he has not been convicted, and he is presumed innocent. The facts about the act come from the complaint and are allegations not yet tested in court. The review used public sources only, and the facts are as of 14 October 2026.
Our earlier case study using the same method, on the co-pilot of flight FZ1073: https://profcheck.co/knowledge/case-study-fz1073-copilot/