· via TechCrunch
Security experts warn nearly everyone in AI is a target for honey-trap spies
TechCrunch reports that amid the Nvidia CEO scandal, security experts warn nearly everyone in the AI industry is exposed to honey-trap espionage, and that AI itself has made such schemes cheaper and easier to run.

What happened
A TechCrunch report published on November 20 argues that the AI industry as a whole is exposed to a form of espionage long associated with politicians, the military and intelligence services: the honey trap. Security experts interviewed for the piece say the sector's most valuable secrets — chip designs, model roadmaps, customer lists and export decisions — now sit inside companies with no tradition of counterintelligence training, making founders, executives and even executive assistants attractive targets.
The report arrives while Nvidia continues to deal with the fallout from allegations against its CEO, Jensen Huang, who has been accused of an affair with Alice Huang, a former executive at a Singapore-based neocloud. She has in turn been accused of working as a Chinese spy and of smuggling billions of dollars of restricted Nvidia chips into China, according to the same reporting.
Not just romance
Aaron Brown, a former senior CIA officer who now runs the software firm Lumbra, told TechCrunch that founders essentially never receive counterintelligence briefings. Honey traps have historically targeted people behind security clearances, in workforces trained to expect an approach from a spy. Today's targets are civilians holding commercially and strategically valuable information, and a CEO is often the shortest path to all of it.
The term suggests romance, but Brown said friendship is actually the oldest variant. Operatives approach targets as investors, customers, recruiters or consultancies — roles whose requests look legitimate on paper and require nobody to fall in love. What the attacker needs, he explained, is a target who feels comfortable talking, and such relationships tend to last longer precisely because nobody is watching for them.
Nor is the risk confined to chief executives. TechCrunch notes that the Wall Street Journal reported OpenAI has begun warning its employees about outsiders trying to get close to them in search of company secrets, and that anyone with access to sensitive information, from board members to assistants, can be targeted. Perry Carpenter, a criminologist and chief deception strategist at security training firm KnowBe4, framed the economics bluntly: for an attacker, charming someone who can influence decisions or make requests nobody wants to question is far easier than fighting through technical defences. And once a target realises they have been compromised, the threat can escalate into blackmail.
AI has cheapened the attack
One irony TechCrunch highlights is that AI itself has lowered the cost of running these operations. Schemes that once demanded months or years of research, persona-building and careful planning can now lean on AI to study a target, fabricate social media profiles and backstories, and even generate fake voice calls. Carpenter said the technology can sustain an entire fictional persona, though the deception itself still plays out through ordinary human relationships.
Recognising the warning signs
The experts offered practical heuristics. Carpenter pointed to unusual patterns, such as coincidental encounters that recur a little too often, persistent curiosity about work and colleagues, requests for secrecy about meeting up, and pressure for introductions or information. A larger red flag is a pattern of requests that keep growing and push past what a target would normally be comfortable with. Brown suggested a simple gut check: would this person still be around if you lost your job tomorrow?
On the defensive side, Carpenter recommends keeping a clear boundary between personal and professional life, and eliminating executive exceptionalism — the assumption that the rules do not apply to people in power. Both stressed company culture: employees need to feel safe raising concerns, challenging unusual requests and escalating possible threats, and a culture where nobody feels able to question the CEO is among the most dangerous. Brown added that founders still need to take risks and pursue opportunities; paranoia does not help, but recognising that you are a target does. Part of what makes these schemes effective, he noted, is that victims often do not know — what feels like a genuine connection, or simply good luck, may be the operation working exactly as designed.
Why it matters
In Brown's assessment, the AI industry acquired national security importance faster than its internal culture adapted to it. Model weights, chip roadmaps, customer relationships and export decisions are now strategic assets, and the people with access to them are the softest layer of the stack. Until companies treat human manipulation as a first-class security concern — alongside phishing, intrusion and insider risk — the cheapest way to steal AI secrets will remain befriending the person who already holds them.
- #ai
- #security
- #espionage
- #nvidia
- #national-security