Jensen Huang, CEO of Nvidia, has acknowledged having an extramarital affair with Alice Huang, a Chinese national whom U.S. intelligence services suspect of having worked for Chinese intelligence. Nvidia has opened an independent investigation, and no theft of industrial secrets has been established at this stage. The case is a reminder that some attacks target trust and human behavior rather than computer vulnerabilities.
Alice Huang, also known as Huang Le, is a former television journalist who became an executive at Megaspeed, a company based in Singapore. According to information initially published by the New York Times on October 29, 2026, she is alleged to have been recruited to gather information about Jensen Huang. In a message to company employees, the executive apologized to his family and stated his willingness to cooperate with the investigation. China denies the espionage allegations.
Megaspeed is also a major Nvidia client in Southeast Asia. According to Associated Press, this company is said to have purchased over $4.8 billion worth of Nvidia processors over two years, amounting to several billion euros. U.S. authorities are also examining suspicions of circumventing trade restrictions imposed on China regarding certain advanced chips. Nvidia’s board of directors says it has no reason to believe the company’s secrets or U.S. national security have been compromised.
It is therefore necessary to distinguish between the relationship acknowledged by Jensen Huang, the suspicions targeting Alice Huang, and the absence, at this stage, of public evidence of a compromise of Nvidia’s secrets.
Social Engineering: An Attack on Human Behavior
Social engineering refers to a set of methods aimed at influencing a person to obtain information, provoke an action, or bypass a security procedure. The attacker does not necessarily seek a computer vulnerability, but an exploitable human reaction. They may rely on trust, fear, curiosity, authority, urgency, greed, professional recognition, or emotional attachment.
An operation can be prepared over several weeks, months, or even years. It often begins with the collection of public information about the target, from LinkedIn, professional publications, conference photographs, or media appearances. This is followed by credible contact, then the building of trust, before an unusual request. The true strength of this method lies in its gradual nature: each stage prepares the next and makes the manipulation less visible.
In intelligence, one technique involves exploiting a person’s romantic or sexual attraction, known as a honey trap. Not all such operations rely on blackmail; some favor trust and the feeling of a privileged relationship. The case of Bernard Boursicot, an employee of the French embassy in Beijing convicted in May 1986 for espionage, provides a historical example.
However, it would be wrong to equate any relationship between a strategic figure and a foreign person with an espionage attempt. The determining criteria lie in the intent to manipulate, the requests made, and the information sought. In Nvidia’s case, only the investigation will be able to establish whether the relationship was exploited for espionage purposes.
Social Networks, Artificial Intelligence: New Grounds
In March 2026, ZATAZ presented techniques for making contact on LinkedIn, TikTok, and personal blogs, in the form of freelance job offers paying a few hundred euros. The requested task often contains no confidential information initially, but the client can later request non-public information. The Direction générale de la sécurité intérieure (DGSI) also warns about these methods, notably flagging fake recruiters or fictitious consulting firms.
Generative artificial intelligence adds new possibilities to these operations. It allows for the rapid production of credible messages in multiple languages and the personalization of initial contact based on public information. Voice synthesis tools can imitate the voice of a known person, and synthetic videos complicate the authentication of interlocutors. The absence of spelling mistakes is therefore no longer a reliable indicator of authenticity.
Effective protection first relies on simple procedures. These include verifying an identity through an independent channel, using already known contact details, and validating any request for money transfer according to a procedure that does not depend solely on the message received. Exercises must be repeated, as scenarios evolve with digital tools, professional habits, and the capabilities of artificial intelligence.
