OpenAI is reportedly warning employees that a flirty stranger could be a foreign spy trying to steal ChatGPT secrets.
The San Francisco-based AI company has cautioned staff about romantic advances, unexpected friendships and seemingly harmless encounters that could expose confidential information.
The warnings, detailed in a 15 October report by The Wall Street Journal, reflect growing concerns about attempts to obtain valuable AI technology. Employees have reportedly received security training to recognise suspicious approaches both inside and outside the workplace.
The concerns have also drawn attention following the arrest of a former OpenAI engineer accused of secretly installing a recording device in a company conference room.
However, authorities have not accused him of working for a foreign intelligence service. The reported warnings also predate his arrest.
Why OpenAI Workers Are Being Warned About Flirty Strangers
OpenAI's reported security guidance addresses a tactic known as a 'honey trap'.
This involves using romantic or intimate attention to build trust and potentially obtain confidential information. An apparently genuine relationship could provide an opportunity to gather sensitive details about AI research.
The Wall Street Journal described a scenario presented to OpenAI employees during security training.
At a San Francisco playground, a stranger repeatedly approached an employee accompanied by their child. The stranger also brought a child and gradually began asking questions about the employee's work. The Journal described the stranger as a spy.
However, the available reporting does not independently verify the person's identity, nationality or intelligence affiliation. It also does not establish whether any confidential information was obtained. The precautions extend beyond romantic encounters.
Employees have been warned about bribes, eavesdropping in cafés and strangers looking at laptop screens. Security measures include clearing sensitive information from whiteboards and avoiding clothing that identifies their employer.
Google and Anthropic have reportedly addressed similar espionage concerns.
FBI Arrest Raises Fresh Questions About OpenAI Security
The warnings have attracted further attention following the arrest of Abdulrahman Alfozan, a 35-year-old former OpenAI engineer, on 9 October 2026.
According to the US Department of Justice, Alfozan allegedly installed a concealed audio recording device inside an executive conference room at OpenAI's San Francisco offices.
Security personnel discovered the device on 30 September. Prosecutors said it was wrapped in toilet paper and inserted into a USB port behind a television. The room was regularly used for confidential discussions about technical research, product plans and business strategy.
Federal investigators alleged that surveillance footage showed Alfozan near the conference room on several occasions. He was not authorised to access the room or attend meetings there.
Security personnel also discovered another recording device inside an unattended bag containing his identification.
OpenAI subsequently dismissed Alfozan and reported the matter to federal authorities. The company said it found no evidence that sensitive model data, source code or customer information had been improperly accessed or extracted.
Alfozan faces a federal charge involving the interception or attempted interception of oral communications.
If convicted, he could receive up to five years in prison and a $250,000 fine. His next court hearing is scheduled for 19 October. He has not been charged with economic espionage or trade secret theft.
Prosecutors have not alleged that he acted for a foreign intelligence agency. The allegations remain unproven, and he is presumed innocent unless convicted.
Why AI Secrets Have Become Valuable Espionage Targets
Concerns about AI-related espionage are not entirely hypothetical.
On 29 January 2026, former Google engineer Linwei Ding was convicted of seven counts of economic espionage and seven counts of stealing trade secrets involving confidential AI technology.
According to the Justice Department, trial evidence showed that Ding stole more than 2,000 pages of confidential information concerning Google's AI computing infrastructure. The material included details about technology used to train and operate advanced AI models.
Prosecutors said Ding obtained the information while pursuing business interests connected to China.
Evidence presented at trial included his involvement with Chinese technology companies and plans to develop AI supercomputing technology using Google's confidential research.
Ding's conviction illustrates the commercial and strategic value of proprietary AI information. However, his case is legally separate from the ongoing OpenAI investigation, in which no economic espionage charge has been announced.
OpenAI's Spy Warnings Spark Online Jokes
The warnings have also prompted jokes about Silicon Valley's dating scene.
Business Insider reported that social media users shared memes suggesting technology workers might now question whether unexpected romantic attention was genuine.
Some joked that being approached by an attractive stranger had become a potential workplace security concern. Others found humour in the idea of AI researchers considering espionage risks before agreeing to a date.
Behind the jokes is a more serious concern about protecting commercially valuable technology. The available reporting does not establish that a romantic honey-trap operation has successfully stolen ChatGPT secrets.
For employees handling confidential AI research, the warnings highlight risks beyond suspicious emails and stolen passwords.
Even an ordinary conversation could become a security concern if someone is deliberately trying to obtain information that was never meant to leave the company.
