White House official says Moonshot AI ‘distills’ the Anthropic model

“We have information that Moonshot AI refined ideas from Fable to develop the K3 model,” Michael Kratsios, Director of the White House office of science and technology policy and science and technology advisor to US President Donald Trump, wrote on X on July 22. “To do that, they developed sophisticated internal platforms, conducted large-scale distillation of US models, and quickly switched between multiple access methods to avoid detection.”

Also according to Kratsios, Moonshot AI quietly purchased the Nvidia GB300 AI chip. Instead of shipping it back to China, the company is said to use it in an Asian country “to train its AI”, violating US export control regulations.

Moonshot AI has not yet commented.

Mr. Kratsios’s accusations were made in the context of Moonshot AI’s Kimi K3 attracting great global attention. Announced on July 17, this is the world’s largest open weight model (publicizing the parameters used in the training process, allowing others to download and modify) with 2.8 trillion parameters, and is expected to be fully released on July 27.

Theo Business InsiderKimi K3 can now process hundreds of pages of text with a single command line, suitable for the task of analyzing long documents and large code bases. Its programming capabilities and agents are competitive with leading models from US companies such as OpenAI and Anthropic, but at a much lower cost.

Experts currently appreciate Kimi K3. Box CEO Aaron Levie called the release of Kimi K3 a “huge win” for companies building on AI platforms, saying it was “really surprising” to see the performance of the open model, as Kimi K3 is only behind Claude Fable 5 and GPT-5.6 Sol in the intelligence index rankings of independent AI evaluation platform Artificial Analysis.

Meanwhile, Nvidia CEO Jensen Huang believes the US has “nothing to fear” or ban China’s open source artificial intelligence model. “The models are excellent,” he told Axios July 22. “We should use excellent open source models.”

Theo The Information On July 22, the US Department of Industry and Security is investigating whether Chinese artificial intelligence companies, including Moonshot, are accessing advanced US AI chips.

A day earlier, Treasury Secretary Scott Bessent said the US would consider whether Chinese artificial intelligence models stole features from US products. “We’ve seen a lot of discussion about open source products emerging and threatening LLMs in the US,” Bessent told Fox Business. “If we determine that foreign models are copying ideas, we have the ability to sanction them.”

Moonshot AI was co-founded by Yang Zhilin, an AI researcher who studied for a doctorate at Carnegie Mellon University (USA), in 2023. According to Reutersthis startup raised more than 2 billion USD in May, bringing its total capital to more than 5.5 billion USD. The company is said to be raising another $2 billion and reaching a valuation of $30 billion.

Moonshot AI was once accused by Anthropic of using “distillation” to train its AI model, along with DeepSeek and MiniMax, through creating 24,000 fake accounts.

In the world of AI, the concept of “distillation” refers to the “transfer of knowledge” from one model to another in a teacher-student fashion. “Distillation is a technique designed to transfer the knowledge of a large pre-trained model (teacher) into a smaller model (student), allowing the student model to achieve performance equivalent to the teacher model,” two scientists Vishal Yadav and Nikhil Pandey once told Forbes. “The technique helps users take advantage of the quality of the LLM, while reducing the cost of inference.”