U.S. Treasury Secretary Scott Bessent attends the 2026 Reagan National Economic Forum at the Ronald Reagan Presidential Library on May 29, 2026 in Simi Valley, California.
Caroline Breman | Bloomberg | Getty Images
U.S. Treasury Secretary Scott Bessent said Tuesday that the Trump administration will investigate whether China’s artificial intelligence models were derived from U.S. models, saying the government does not support “intellectual property theft.”
China’s open weight model is gaining traction against major products from US companies. OpenAI and humanraising concerns from technology executives and government officials about the sustainability of America’s lead in the AI race. Earlier this month, Chinese startup Moonshot AI released a model called Kimi K3 that outperforms these companies on some industry benchmarks. Openweight refers to a model in which the underlying code and training data remain private, while the final trained parameters are publicly released for download.
“Especially when we see foreign models stealing from our great companies, we have the ability to sanction models for this theft,” Bessent said Tuesday on Fox Business’ “Morning with Maria.”
Bessent said the technical term for this theft is distillation, an AI training method that uses the output from an existing, more powerful model to build a smaller, less capable model. Last month, Anthropic sent a letter to the U.S. Senate Banking, Housing and Urban Affairs Committee alleging that Chinese tech company Alibaba had launched “the largest distillation attack ever known” against the company.
The United States and China are scheduled to hold talks on AI in September, Reuters reported on Tuesday. Bessent will represent the United States during the talks, the report said.
“We’ve found watermarks from the U.S. large-scale language model on many Chinese models, and this is unacceptable,” Bessent said Tuesday. “We’ll be looking at that in the coming days and weeks.”
However, both Anthropic and OpenAI have faced accusations of theft in recent years.
In September, Anthropic agreed to pay $1.5 billion to settle a class-action lawsuit with a group of authors who say the company illegally downloaded books from a pirate database. The New York Times sued OpenAI and Microsoft in 2023, accusing them of infringing copyright by using their intellectual property to train models.
