OpenAI is reportedly advising its employees to be cautious of potential foreign intelligence operatives who might attempt to exploit them to access confidential company information. This warning, highlighted in a significant report, includes caution against “honey pot” tactics specifically designed for espionage, as indicated by unnamed sources, including an incident where a Saudi engineer was found placing secret recording equipment in company conference rooms.
The report recounts an unsettling scenario in which a stranger approached an OpenAI worker at a playground in San Francisco while they were with their child. The stranger engaged in light conversation initially but then insidiously directed the chat towards the employee’s work at OpenAI. It later turned out that the individual was a spy, according to the Journal.
Through training sessions focused on the risks of “honey pot” operations, staff have been alerted about how foreign agents might use flirtation to gain access to sensitive information. An illustrative example shared with employees involved a US diplomat in Russia who was targeted by foreign spies dispatching a blonde woman to his hotel room accompanied by Champagne.
When the diplomat declined her advances, the spying agency succeeded in sending a redhead and even a young man as further attempts to entice him, as sources relayed to the Journal.
OpenAI’s security professionals have laid out additional tactics that spies might employ, such as bribing employees or listening in on conversations in public venues like cafes. Consequently, staff have been instructed to take protective measures, such as using screen protectors for their devices and erasing any notes from whiteboards after meetings. They’re also advised against showcasing company logos in public settings.
Interestingly, other AI firms, including Anthropic and Google, are reportedly issuing similar security advisories to their personnel.
While the idea of “honey pot” schemes may sound far-fetched to some, they are indeed recognized threats in business and political contexts. A notable case involves former Democratic Representative Eric Swalwell, who admitted to engaging in relations with a suspected Chinese agent named Christine Fang, also referred to as “Fang Fang,” as documented in files recently released by the White House Transparency Task Force.
The accused Saudi engineer, Abdulrahman Alfozan, aged 35, was apprehended last Friday and faces charges for allegedly using devices to intercept oral communications. The U.S. Justice Department disclosed that OpenAI had found an audio recording device cleverly hidden in a USB port behind a television in one of their executive conference rooms, concealed within toilet paper to avoid detection.
After the arrest, FBI Director Kash Patel emphasized the importance of protecting sensitive AI development data, labeling it critical for maintaining the nation’s economic strength and security, which the FBI considers a top counterintelligence priority.
Alfozan could face a sentence of up to five years in prison and a substantial fine of $250,000.
