China’s Moonshot Reportedly Stole U.S. AI Tech for Kimi K3

White House AI adviser Michael Kratsios accused Chinese startup Moonshot AI of secretly copying Anthropic’s Claude Fable 5. He said the copies helped build its new Kimi K3 model.

Kratsios made the claim in a Wednesday post on X. He said Moonshot built a special platform for large-scale distillation of US models. The firm allegedly kept switching access routes to avoid getting caught.

White House Draws a Line on Moonshot AI Distillation

The technique at the center of the row is called distillation. It trains a new AI model on the answers of a stronger one. Done openly, it is legal and common. Kratsios said the covert, industrial-scale version is theft.

“We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model… However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable,” Michael Kratsios, Director of the White House Office of Science and Technology Policy stated.

Anthropic has said the same thing. In a February report, the company traced over 3.4 million Claude exchanges to Moonshot. The traffic came from hundreds of fake accounts. Account records even matched the public profiles of senior Moonshot staff.

Anthropic said the campaign targeted Claude’s reasoning, coding, and vision skills.

Kratsios also said Moonshot got hold of NVIDIA GB300 servers and used them in Thailand. These are NVIDIA’s top AI chips, and US rules block their sale to China. Washington tightened its AI chip export clampdown in May. In June, a Super Micro office raid in Taiwan targeted alleged chip smuggling into China.

Kimi K3 Deepens the US-China AI Fight

Moonshot launched Kimi K3 on July 16. At 2.8 trillion parameters, it is the biggest open-weight model ever released. Demand was so strong that Moonshot had to pause new Kimi subscriptions within 48 hours. The full model weights go public by July 27.

That date matters. Anthropic warns that copied models lose their safety guardrails. Once open-sourced, they can spread beyond anyone’s control.

THIS IS ABSOLUTELY 100% THE WRONG WAY TO DEAL WITH CHINA’S AI LEAD.“OpenAI and Anthropic executives are sounding the alarm about the rise of cheap AI, particularly powerful new models produced in China, suggesting they will lead to a ‘dystopian’ AI future and present… https://t.co/pg3nCVkLuK pic.twitter.com/61aNsgeh7B

— Brian Roemmele (@BrianRoemmele) July 21, 2026

This is not the first such charge. In January 2025, White House AI czar David Sacks accused DeepSeek of copying OpenAI’s models. DeepSeek denied it. Beijing called Anthropic’s February findings groundless.

Kratsios announced no penalties, and Moonshot has stayed silent. The US still holds a 23x spending lead in AI, yet Chinese models keep gaining ground. The next move now sits with Washington.