U.S. Treasury Secretary Bessent: Sanctions may be imposed if Chinese AI models are confirmed to have "stealed" U.S. technology. U.S. Treasury Secretary Scott Bessant said on Tuesday that the Trump administration will investigate whether China's artificial intelligence models obtain results from U.S. models through distillation technology, and stressed that the U.S. government will never support intellectual property theft.

U.S. Treasury Secretary Bessent: Sanctions may be imposed if Chinese AI models are confirmed to have "stealed" U.S. technology. U.S. Treasury Secretary Scott Bessant said on Tuesday that the Trump administration will investigate whether China's artificial intelligence models obtain results from U.S. models through distillation technology, and stressed that the U.S. government will never support intellectual property theft. Currently, China's open source models are gradually catching up with core products of leading American companies such as OpenAI and Anthropic in terms of performance. This has triggered concerns among American technology executives and government officials about the sustainability of the United States' leading position in the global race for artificial intelligence. Earlier this month, Chinese startup Dark Side of the Moon released a new model called Kimi K3, which has outperformed the aforementioned American company's products in some industry benchmarks. In an interview with Fox Business Channel, Bessant pointed out that if overseas models are found to be stealing the results of outstanding American companies, the United States is fully capable of imposing sanctions against such theft. Bessant explained that this behavior is technically known as "model distillation", which is to use the output of an existing stronger model to train and build a smaller, less powerful model. Last month, Anthropic sent a letter to the U.S. Senate Committee on Banking, Housing and Urban Affairs, accusing the Chinese technology company Alibaba of carrying out the largest distillation attack to date. According to a Reuters report on Tuesday, the United States and China plan to hold relevant talks on artificial intelligence issues in September this year, at which time Bessant will attend the discussions on behalf of the United States. Bessant said in the interview that the United States has discovered digital watermarks of large American language models in many Chinese models, which is unacceptable and that the United States will conduct an in-depth investigation into this in the coming days or weeks. However, Anthropic and OpenAI themselves have faced similar copyright infringement accusations in recent years. Last September, Anthropic agreed to pay $1.5 billion to settle a class-action lawsuit filed by a group of writers who accused the company of illegally downloading and using books from a pirated database for training. The New York Times also sued OpenAI and Microsoft in 2023, accusing them of using the newspaper’s intellectual property without authorization to train AI models. via cnBeta.COM - Chinese industry information station (author: source: cnBeta.COM)