US Government Alleges Chinese AI Firm Moonshot AI Illegally Used Anthropic’s Model to Develop K3

The United States government has formally accused Chinese artificial intelligence startup Moonshot AI of misappropriating Anthropic’s advanced large language model, Fable, to develop its recently launched K3 AI model. This development was disclosed by senior US officials, including Michael Kratsios, the White House’s chief technology officer, and Treasury Secretary Scott Bessent.

Kratsios stated on the social media platform X that the US possesses information indicating Moonshot AI engaged in "large-scale, covert industrial distillation" of Anthropic’s Fable model to create its K3 system. He emphasized that such unauthorized use of proprietary US technology undermines American research and is unacceptable.

In a separate post on X, Treasury Secretary Bessent announced that he is considering placing Moonshot AI on a US trade blacklist and imposing sanctions as a response to these allegations. These actions reflect growing concerns over intellectual property theft and technology transfer risks amid intensifying competition in AI development between the US and China.

The accusations come shortly after Moonshot AI unveiled Kimi K3, a 2.8 trillion-parameter open-weight AI model that the company claims approaches the performance of Anthropic’s Fable. This launch underscores the rapid advancement of China’s AI capabilities and its narrowing gap with leading US developers.

Anthropic has previously reported that three Chinese AI companies—DeepSeek, Moonshot AI, and MiniMax—violated its terms of service by using its Claude model to improve their own AI systems. According to Anthropic, these companies created over 16 million interactions through approximately 24,000 fake accounts, breaching regional access restrictions.

Model distillation, the process at the center of these allegations, involves training smaller AI models using outputs generated by larger, more capable models. This technique can reduce development costs but raises intellectual property concerns when done without authorization.

Further complicating the situation, Kratsios alleged that Moonshot AI acquired servers equipped with Nvidia’s GB300 AI chips and deployed them in Thailand, likely to train its AI models. This claim echoes earlier accusations from a senior Trump administration official that Chinese AI startup DeepSeek used Nvidia’s advanced AI chips in violation of US export controls.

The Chinese embassy in Washington has rejected the US allegations as "entirely unfounded." Embassy spokesperson Liu Chang stated that China respects intellectual property protections and urged US officials to "respect the facts, discard prejudice and stop smearing and discrediting China’s achievements in the development of its artificial intelligence industry."