A senior White House advisor has publicly accused Chinese artificial intelligence startup Moonshot AI of illicitly using a proprietary model from American firm Anthropic to develop its new Kimi K3 AI. Michael Kratsios, a science and technology advisor to President Donald Trump, detailed the allegations in a recent social media post, claiming Moonshot AI engaged in "large-scale distillation" of Anthropic's Claude Fable 5 model.
Allegations of Distillation and Restricted Hardware Use
Distillation is a technique where a smaller or less capable AI model learns from the responses and behavior of a more advanced, larger model. While not inherently illegal, using a proprietary model for distillation without explicit permission typically violates the terms of service set by companies like Anthropic and OpenAI.
Kratsios further alleged that Moonshot AI utilized a "sophisticated internal platform" to conduct this distillation against US models, enabling them to bypass detection by quickly switching access methods. He also claimed the Chinese company obtained and used export-controlled, high-end Nvidia AI servers, hardware subject to US restrictions aimed at limiting China's access to powerful computing for AI development.
"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 stated.
Moonshot AI launched its Kimi K3 model last week, asserting its performance rivals Anthropic's Fable 5 and surpasses several other flagship US AI models.
Broader US Concerns and China's Response
These accusations align with growing concerns from the US government regarding potential intellectual property theft in the AI sector. Just a day prior to Kratsios's claims, US Treasury Secretary Scott Bessent announced that the US is investigating whether Chinese AI models have stolen capabilities from American counterparts. Bessent warned that sanctions could be imposed on Chinese companies found to be engaged in "industrial-scale distillation attacks" that constitute IP theft.
In response to the allegations, Liu Chang, a spokesperson for the Chinese embassy in Washington, dismissed the claims as "entirely unfounded." Liu 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."
The dispute highlights escalating tensions and a broader geopolitical competition in artificial intelligence, encompassing both technological advancements and strategic national security interests.