US accuses China’s Moonshot AI of using Anthropic model to develop K3

Separately, US Treasury Secretary Scott Bessent said in a post on X that he was considering adding Moonshot AI to a US trade blacklist and imposing sanctions.The Donald Trump administration has accused Chinese artificial intelligence (AI) startup Moonshot AI of using Anthropic’s most advanced large language model, Fable, to help develop its recently launched K3 model, according to Reuters.Michael Kratsios, the White House’s top technology official, said the US government had information suggesting Moonshot AI used Anthropic’s model during the development of K3.“We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model,” Kratsios said in a post on X.“Large-scale, covert industrial distillation aimed at stealing proprietary US technology and undermining American research is unacceptable,” he added.Separately, US Treasury Secretary Scott Bessent said in a post on X that he was considering adding Moonshot AI to a US trade blacklist and imposing sanctions.The allegations are expected to further heighten technology tensions between Washington and Beijing as AI companies in both countries race to develop increasingly advanced models with commercial and strategic applications.A spokesperson for the Chinese embassy in Washington rejected the allegations, describing Kratsios’ comments as “entirely unfounded”.“China respects intellectual property protections,” embassy spokesperson Liu Chang said, adding that US officials should “respect the facts, discard prejudice and stop smearing and discrediting China’s achievements in the development of its artificial intelligence industry”.Anthropic public policy executive Sarah Heck said in a post on X that Chinese theft of US AI models “creates serious national security risks for the United States.”Nvidia chip claims add to tensionsKratsios also alleged that Moonshot AI had acquired servers equipped with Nvidia’s GB300 AI chips and deployed them in Thailand, where they were “likely to train its AI models”.Reuters noted that a senior Trump administration official had previously alleged in February that Chinese AI startup DeepSeek trained its latest model using Nvidia’s most advanced AI chips, potentially violating US export controls.Model distillation refers to the process of training smaller AI models using outputs generated by larger, more capable models to reduce development costs.Anthropic said in February that three Chinese AI companies—DeepSeek, Moonshot AI and MiniMax—used its Claude model to improve their own AI systems by creating more than 16 million interactions through approximately 24,000 fake accounts, violating its terms of service and regional access restrictions.The allegations come days after Moonshot AI unveiled Kimi K3, a 2.8 trillion-parameter open-weight AI model that the company said approaches the performance of Anthropic’s Fable model.

The launch highlighted the rapid progress of China’s AI industry as it narrows the gap with leading US developers.