US Treasury vs. Nvidia: Jensen Huang Pushes Back on Sanctions Against Chinese AI

On July 21, 2026, US Treasury Secretary Scott Bessent said Washington will begin screening Chinese open-source AI models for stolen intellectual property. The same day, Nvidia CEO Jensen Huang publicly objected, calling the logic backwards.

What the US Treasury said

Scott Bessent, the US Treasury Secretary, warned China in a televised interview on July 21, 2026 that American authorities will closely scrutinize Chinese open-weight models for stolen intellectual property. Bloomberg was the first to report the administration's plans.

The targets are open-source models, which are published with open weights and have become real competitors to Western systems over the past year. Bessent's wording ties their spread directly to the "theft" of technology from American companies.

• Date of the statement: July 21, 2026 (a Tuesday)
• Who: Scott Bessent, US Treasury Secretary
• The gist: screening Chinese open-source AI models for stolen intellectual property
• Who revealed the plans: Bloomberg
• Opponent within the industry: Jensen Huang, CEO of Nvidia

Why Jensen Huang objects

Jensen Huang, Nvidia's CEO, took the opposite position on July 21, 2026. In his assessment, trying to punish Chinese models for "theft" turns the problem on its head.

His words carry weight because of Nvidia's position. The company is the largest supplier of the GPUs on which most modern AI models are trained, including the very Chinese systems the Treasury is talking about. That is why a split between Nvidia's chief and a US financial regulator on the same day is more than an ordinary spat. It signals a rift between the industry and policymakers.

"Nvidia's

Jensen Huang says that's backwards" is how The Next Web headlines the public disagreement between Nvidia's chief and the US Treasury.

What's behind the dispute

The "theft" accusations sit atop a years-long conflict over China's access to advanced AI technology. The US has already restricted exports of powerful chips, and Jensen Huang has repeatedly argued that tough restrictions hurt American manufacturers and only push China toward its own technological independence.

The new twist on July 21, 2026 moves the fight from hardware to the models themselves. Where the debate used to be about chip sales, Washington is now, according to Bloomberg, aiming at the software side: open models that anyone can download and run anywhere. Sanctioning code is far harder than sanctioning physical shipments, and Nvidia's stance suggests that is exactly what makes the approach contentious.

What this means

The Bessent-Huang dispute exposes the central question of the US-China AI race: are open Chinese models a security threat or an inevitable product of global competition? Whose logic prevails will determine whether open-source models come under a sanctions regime for the first time, and how Nvidia, whose business depends on global demand for AI chips, will react.

Frequently asked questions

Why does the US want to screen Chinese AI models?

According to Treasury Secretary Scott Bessent (July 21, 2026), US authorities intend to examine Chinese open-source models for intellectual property stolen from American companies. Bloomberg was the first to report the administration's plans.

Why does Nvidia's chief disagree with Washington?

Jensen Huang, Nvidia's CEO, considers this approach backwards. His company is the world's largest supplier of AI chips, and restrictions targeting Chinese models affect the entire global market in which Nvidia operates.