The Nvidia investigation has made corporate espionage in technology visible again. It is worth knowing how old the problem is, because the history changes how you read the current case.
France, IBM, and 1987
In the late 1980s the PC revolution was under way and American companies were running circles around Compagnie des Machines Bull, the struggling French computer firm.
France decided to find out how. Between 1987 and 1989, the Direction Générale de la Sécurité Extérieure — France's equivalent of the CIA — planted moles inside the offices of Bull's biggest American competitors, IBM and Texas Instruments, where they passed trade secrets back to Bull.
The FBI and CIA eventually uncovered the scheme and the employees were fired.
Four decades on
The problem has grown, and so have the stakes. Nvidia, the world's largest and most important AI chipmaker, is investigating reports that it was the victim of a foreign surveillance effort; its chief executive has separately acknowledged an extramarital affair. Neither the company nor Huang has connected the two.
And it is not one company. OpenAI, Google, Apple and other high-profile technology firms have reported foreign and domestic spying attempts in recent weeks and months.
"Corporate espionage has always been a big challenge, but today it's especially so in areas where US firms are building technologies that are foundational to economic, military and other forms of power," said Matthew Waxman, a cybersecurity law professor at Columbia University.
The line worth sitting with
CNN's assessment is that Big Tech's spy problem is almost certainly a lot bigger than anyone realises.
That follows from the structure of the thing rather than from any estimate. Every case in the public record is a case that failed — the mole who was caught, the breach that was noticed, the approach that was reported. Successful espionage produces no event to count. So the reported total is a measure of detection, not of activity, and the gap between them is unknowable by definition.
Why AI intensified it
As information became one of the world's most valuable commodities, technology companies' secrets became the gold in the vault. AI has sharpened that considerably: the value is concentrated in model weights, training methods and chip designs — things that can be carried out of a building on a drive, unlike a factory or a drug pipeline.
That is also why export controls and espionage have converged into one story. When the restricted item is a design rather than an object, enforcing a border and preventing a leak become the same problem.
For Bangladesh, the practical consequence is already familiar from the Nvidia coverage on this site: the rules governing which chips reach which buyers are being written in a security frame, not a trade one. That tends to make them tighter, slower to change, and harder to argue with.
What companies do about it, and who pays
The standard response is an insider-risk programme: background checks, access logging, monitoring of file transfers and unusual activity, and restrictions on who can see what.
That has a cost nobody in the press release mentions. Those controls fall on every employee, almost none of whom are spies. The result is a workplace where ordinary engineers are monitored continuously because a very small number of people in the industry are not trustworthy — and the tighter the espionage environment gets, the more surveillance of staff becomes normal.
