Experts warn AI industry staff are exposed to honey-trap espionage schemes
A former CIA officer and a deception specialist say founders, executives and assistants at AI firms are targets, and that AI tools now make such schemes cheaper and faster to run.
· 2 min read
- Security experts told TechCrunch that people across the AI industry, not only CEOs, can be targeted by honey-trap schemes.
- AI tools can now research targets and build fake profiles, backgrounds and voice calls, which lowers the cost of such operations.
- The experts advise watching for unusual patterns, setting personal and professional boundaries, and ending special treatment for executives.
People working in the artificial intelligence industry face a growing risk of honey-trap schemes, according to security specialists quoted by TechCrunch. The warning comes as the fallout continues from allegations involving Nvidia chief executive Jensen Huang.
TechCrunch reported that Huang has been accused of falling for such a scheme through a relationship with Alice Huang, a former executive at a Singapore-based neocloud company. She has been accused of being a Chinese spy and of smuggling restricted Nvidia chips worth billions of dollars into China. These are allegations, and the TechCrunch report presents them as such.
Aaron Brown, a former senior CIA officer who now runs the software company Lumbra, said founders are rarely briefed on counterintelligence. He noted that honey traps have traditionally been linked to politicians, the military and intelligence services, whose staff are trained to expect approaches. Many of today's most valuable secrets, he argued, now sit inside companies that build chips, models and infrastructure, which makes chief executives an efficient route to roadmaps, export decisions and customer lists.
The risk is not limited to the top of the organization. TechCrunch cited a Wall Street Journal report that OpenAI has begun warning its employees about outsiders who try to befriend them in order to obtain company information. Board members and executive assistants are among those with access that could attract attackers.
The experts also stressed that these schemes are often not romantic. Brown said friendship is the oldest approach, and that attackers may pose as investors, customers, recruiters or consultants. Their requests look legitimate, and the goal is to make the target comfortable talking. Once a target realizes they have been compromised, blackmail becomes possible.
AI itself is making the tactic cheaper. Perry Carpenter, chief deception strategist at KnowBe4, said that planning such an operation once took months or years of research and persona building. Now AI can help study a target and produce fake social profiles, invented backgrounds and even synthetic voice calls. Carpenter added that the technology sets the stage, while the deception still plays out through human relationships.
For protection, Carpenter suggested watching for coincidental encounters that happen too often, constant curiosity about work and colleagues, requests for secrecy and pressure for introductions or information. Requests that grow steadily bigger are a further warning sign. Brown proposed a simple test: would this person still be around if you lost your job tomorrow?
Carpenter also recommended a clear line between personal and professional life, and the end of what he called executive exceptionalism inside companies. He warned that a culture in which nobody feels able to question the CEO is among the most dangerous for security. Brown said founders must still take risks and explore opportunities, adding that paranoia does not help, but recognizing that you are a target does.