US Threatens Chinese AI with Sanctions Over IP Theft

Treasury Secretary Scott Bessent announced on Tuesday that the United States intends to scrutinize open-source artificial intelligence models originating from China for evidence of intellectual property theft. Should such theft be confirmed, sanctions against Chinese AI companies would be a potential consequence.

Speaking on Fox Business, Bessent remarked, “We’ve seen a lot of talk about open source models coming and threatening the large language models in the US.” He clarified the administration's position: “This administration supports open source models, but what we do not support is IP theft. If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft.”

Bloomberg was the first to report on Secretary Bessent’s comments.

This declaration emerges amidst the increasing sophistication and popularity of Chinese AI models, notably Moonshot AI’s Kimi K3. Their advancement poses a competitive challenge to the business frameworks of leading American AI firms like OpenAI and Anthropic, potentially impacting their capacity to secure further investment for the development of cutting-edge models.

Earlier in the week, Axios reported that the Trump administration was contemplating a comprehensive prohibition on Chinese open-source models, although this report has since faced refutation from various sources.

For months, AI companies have voiced concerns regarding efforts by foreign entities to replicate their AI technologies and subsequently release them as open-source. In response, the White House committed in April to collaborate closely with AI firms to counteract such illicit appropriation.

The imposition of U.S. sanctions on Chinese AI models would constitute another measure in the government’s expanding toolkit designed to maintain America’s leadership in the global AI race. Following previous actions such as restricting China’s access to advanced semiconductors and tightening export controls, Washington is now signaling a potential direct targeting of AI models themselves. This move could signify a substantial escalation in the technological rivalry between pioneering U.S. laboratories and Chinese open-source alternatives.

A technique known as model distillation allows for the transfer of some capabilities from a larger AI model into a smaller, more efficiently runnable system. However, there is no universal consensus that distilling another company’s model inherently constitutes intellectual property theft.

Microsoft CEO Satya Nadella voiced criticism this month regarding the assumption made by some large labs that distillation equates to theft. He stated, “While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation.”

The training methodologies employed by AI labs continue to present considerable legal risks. This week, Anthropic received approval to begin issuing payments to authors as part of its $1.5 billion settlement. The settlement followed a judicial ruling that the company had unlawfully downloaded and stored millions of copyrighted books to train its AI systems.

Moreover, some industry experts contend that model distillation alone is not the primary driver behind China’s progress in catching up with U.S. AI companies.

Clem Delangue, CEO of Hugging Face, commented on a recent episode of TechCrunch’s Equity podcast, “We know distillation to be a very small factor in the ability to create good models, and it’s a practice that everyone is doing, including companies in the U.S.” He further elaborated, “If it were easy just to do distillation to get good at building AI models, there would be many other countries, including in the U.S., with much better open source AI. The reality is they have really, really good research teams in China…taking a much more open and collaborative approach to AI than in the U.S.”