“We’ve seen a lot of talk about open source models coming and threatening the large language models in the US,” Bessent said. “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.”
Bessent pointed to model distillation, a technique that uses outputs from a larger, more capable model to train a smaller one, as the practice under scrutiny. He said the administration is evaluating whether some Chinese AI models contain evidence of being derived from American systems.
“We are finding watermarks of our U.S. large language models on many of the Chinese models, and that’s unacceptable,” Bessent said Tuesday. “We’re going to be looking at that in the coming days or weeks.”
The comments follow growing attention on China's AI sector, including Moonshot AI's recently released Kimi K3 model, which has outperformed leading models from OpenAI and Anthropic on some industry benchmarks. The rapid progress of Chinese open-weight models has prompted concerns among technology executives and U.S. officials about maintaining the country's position in AI development.
Anthropic has previously accused Chinese companies of using distillation to copy its technology. Last month, the company told the U.S. Senate Committee on Banking, Housing, and Urban Affairs that Alibaba had carried out what it described as “the largest known distillation attack” against Anthropic.
According to Reuters, the United States and China are planning AI-focused talks in September, with Bessent expected to represent the U.S. during the discussions.
The possibility of sanctions would add another measure to the U.S. government's efforts to protect domestic AI technology. Washington has already imposed restrictions on advanced AI chip exports to China, and Bessent's comments suggest AI models themselves could become a future focus of enforcement if intellectual property violations are established.
At the same time, the use of model distillation remains a subject of debate within the AI industry. Some executives argue the technique is a common practice and does not necessarily constitute theft. Microsoft CEO Satya Nadella recently questioned the industry's position, saying, “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.”
American AI companies also continue to face legal challenges over their own training practices. Anthropic recently received final approval for a $1.5 billion settlement in a copyright lawsuit alleging it illegally downloaded and stored copyrighted books to train its AI models. OpenAI and Microsoft have also been sued by The New York Times over claims that copyrighted material was used to develop their AI systems.
This analysis is based on reporting from CNBC.
Image courtesy of Unsplash.
This article was generated with AI assistance and reviewed for accuracy and quality.
