US claims China's Moonshot AI used Anthropic's Fable 5 to build Kimi K3 AI model

President Donald Trump's Science and Technology adviser, Michael Kratsios, accused Chinese AI startup Moonshot AI of conducting a “large-scale” distillation campaign for its new Kimi K3 AI model. The startup released the model last week, claiming that it performs similarly to Anthropic’s Fable 5 model, and even better than a few US flagship AI models.

What is Distillation?

Distillation is an AI training technique where a smaller or less capable model is trained from a more advanced model and uses its responses as training data. While distillation is a common process among AI companies, if done without permission, it can violate policies, and Anthropic and OpenAI generally prohibit using their proprietary models to train competing AI systems without permission.

Allegations on Moonshot AI for Kimi K3 model

Michael Kratsios recently shared an X post accusing Moonshot AI of training its Kimi K3 model on Anthropic’s Claude Fable 5. The White House official also accused Moonshot of obtaining access to export-controlled, high-end Nvidia AI servers, hardware that is subject to US restrictions to train its models.

“We have information ‌that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model. To do this, they developed a sophisticated internal platform to conduct large-scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection.” Kratsios said.

Kratsios alleged that Moonshot likely relied on Nvidia's GB300 Grace Blackwell-powered servers to train its models. In 2022, Washington restricted exports of Nvidia's most advanced AI chips to China to limit its access to powerful computing power needed for AI development. According to a BBC report, governments in several countries have increased efforts to stop the illegal smuggling of these restricted chips into China.

Kratsios' accusations came just a day after US Treasury Secretary Scott Bessent said that the US is investigating whether Chinese AI models had stolen the capabilities of American AI models.

Bessent also warned that the US could impose sanctions on Chinese companies if it determines they engaged in “industrial-scale distillation attacks” in a way that constitutes intellectual property theft.

Reuters also quoted Spokesperson Liu Chang of the Chinese embassy in Washington as saying that the allegations are “entirely unfounded” and “Relevant individuals in the United States should respect the facts, discard prejudice and stop smearing and discrediting China’s achievements in the development ​of its artificial intelligence industry.”