Despite their apparent importance to national security, AI labs are not subject to the same rigorous vetting procedures as aerospace and defense companies.
“The global race for AI has put Western technology companies on the front lines of state-sponsored espionage,” says David Kilby, a former senior government security official now at corporate intelligence firm Strider Technologies. “They need a clearer idea of who they are hiring and who might try to touch their employees once they arrive.”
Some positions within technology companies require a higher level of verification. An OpenAI job description for working on the company’s national security programs requires a top-secret security clearance and possibly a polygraph test.
The company has also hired “inside risk investigators,” which involve detecting and analyzing unusual employee behavior. A job description for a similar role at Anthropic says the company uses its own AI models to try to eliminate spies.
OpenAI has trained its employees to spot attempts to steal its secrets. According to the Wall Street Journal, he warned that spies might try to seduce employees, befriend them in public spaces or eavesdrop on conversations in cafes. Staff are also advised not to wear company-branded clothing in public.
Some meeting rooms are off-limits to most employees (suspicions were reportedly raised around Alfozan when he was seen loitering around a conference room he was not allowed to enter), with some requiring fingerprints to be taken.
However, companies have also been criticized for not taking the threat seriously. Leopold Aschenbrenner, a former researcher, claimed that he was actually fired for raising security concerns.
He claimed that after he raised the threat of Chinese espionage, the company’s HR department accused him of racism (OpenAI claims Aschenbrenner was actually fired for disclosing information, although it disputes that the information was confidential).
Gn bussni
