According to a controversial report, OpenAI has warned its employees to be wary of foreign spies who might try to seduce them into stealing company secrets.
So-called “honeypot” systems are among the dangers OpenAI flagged to warn its engineers against espionage, the Wall Street Journal reported, citing anonymous sources. The risks became even more urgent this week after a Saudi engineer was caught installing secret recording devices in executive conference rooms.
The report cites an incident in which a stranger approached an OpenAI employee who was with his child at a San Francisco playground and began chatting – before shifting the conversation to his work at the AI company. The stranger was actually a spy, according to the Journal.
During training related to “honeypot” programs, OpenAI employees are warned that foreign agents could attempt to use gender to get closer to a target and steal secrets. To illustrate the risk, OpenAI told collaborators about a US diplomat based in Russia who was targeted by foreign spies who sent a blonde woman to his hotel room with champagne.
When the diplomat refused to let the woman in, foreign intelligence sent a red-haired woman and then a young man to try to lure her out, sources told the Journal.
OpenAI security experts also described other possible methods used by spies, including offering bribes or eavesdropping on employees in public spaces like cafes, the report said.
Employees are specifically asked to take steps to protect themselves, including using screen protectors to hide their phones and computers in public and erasing any information on whiteboards after meetings. They are also asked to avoid wearing clothing bearing the company logo in public.
Other companies in the high-stakes AI race, including Anthropic and Google, have reportedly given similar warnings to their employees.
Honeypot projects may seem far-fetched, but they pose a well-known threat to business and political figures.
One of those cases allegedly ensnared former Democratic Rep. Eric Swalwell, who admitted to having “physical relations” with suspected Chinese spy Christine Fang, also known as “Fang Fang,” according to investigative documents released by the White House Transparency Task Force in August.
The Post has contacted OpenAI for comment.
The Saudi engineer who installed a listening device at OpenAI headquarters has been identified as 35-year-old Abdulrahman Alfozan. Federal authorities arrested him last Friday and charged him with “using or attempting to use an electronic, mechanical or other device to intercept oral communications.”
The Justice Department said OpenAI “discovered a surreptitious audio recording device (Recording Device 1) inserted into a USB port on the back of a television in an executive conference room inside one of their buildings in San Francisco.” The device “had been concealed in toilet paper” in an attempt to avoid suspicion.
After the arrest, FBI Director Kash Patel said protecting confidential information about AI development was “critical to safeguarding our nation’s economic strength and national security, and the FBI views this mission as one of our highest counterintelligence priorities.”
Alfozan faces up to five years in prison and a $250,000 fine.
Gn bussni
