NVIDIA CEO Jensen Huang stated that the “distillation” of AI models—i.e., training a model using the outputs of other models—is considered “competition,” although the White House refers to it as theft.

In the competition between the US and China for dominance in artificial intelligence, distillation has become a key point of contention in their tense relations. US officials accuse a Chinese AI company of using distillation technology.

U.S. Treasury Secretary Scott Bessent described it as "thiefery" in July and threatened to impose sanctions on overseas companies that use this method to extract capabilities from U.S.-developed models.

However, when asked whether distillation is "not robbery," Jensen Huang said in an interview with the CNBC program " Squawk Box " on Monday: "That's what competition is all about."

He said, "You're completely free to test other people's products to your heart's content." He also added that sometimes NVIDIA's products are "disassembled down to their bare bones" by companies in an attempt to understand how they work.

He said, "Of course, I would prefer if they didn't do that. I certainly wish no one would learn from our products so that we could move forward smoothly. But to be honest, competition will make everything better."

Jensen Huang also said, "If you don't like that, if you don't want others to use your product, then what you need to do is to understand your customers, and then shut down that service."

The White House did not immediately respond to the request for a comment from CNBC.

Earlier this month, the US Cybersecurity and Infrastructure Security Agency accused a Chinese AI company of conducting "industrial-scale knowledge distillation operations," claiming that it violated the terms of use of American companies.

Anthropic stated earlier this month that it had discovered that Alibaba, which develops the Tongyi Qianwen (Qwen) model family, as well as DeepSeek, were involved in "illegal distillation."

The Chinese side denied these accusations.