Moonshot AI Faces Allegations of Extracting Anthropic's Technology for K3 Model
Washington threatens sanctions and export controls against Chinese companies engaged in large-scale, clandestine AI model distillation operations.
The White House has leveled accusations against Moonshot AI, claiming the company engaged in distillation of Anthropic's Fable AI model as part of its development process for Kimi K3, a model that debuted last week.
Michael Kratsios, Director of the White House Office of Science and Technology Policy, took to X on Wednesday to make public allegations that the Chinese artificial intelligence company built an internal infrastructure specifically designed to distill American AI models on a massive scale, employing techniques intended to avoid detection.
"Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem," he said. "However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable."
The Kimi K3 model has risen to prominence as among China's most advanced AI systems, amplifying concerns in Washington that Chinese developers are secretly leveraging American models to fast-track their own artificial intelligence development.
Nevertheless, certain researchers in the AI community have raised doubts about assertions that Anthropic's most recent AI model served as training material for Kimi K3.
Fable 5 from Anthropic underwent a re-release on July 1 following its rapid withdrawal from availability due to United States export control regulations, whereas Kimi K3 made its debut on July 16, leaving only a brief timeframe during which distillation attacks could have taken place.
"There are only 15 days between fable 5 ban removal and kimi K3 release," said Elie Bakouch, a researcher at AI startup Prime Intellect.
"I don't think claiming that K3's performance comes from fable distillation (even if they did it) makes sense technically."
On Friday, Dean Ball, OpenAI's head of strategic futures, stated he didn't believe the K3 model's performance could be "explained away by distillation or anything like that."
Scott Bessent, US Treasury Secretary, issued a warning that such large-scale distillation attacks might trigger sanctions along with additional restrictive measures.
"We support open-source AI and the innovation it unlocks. But open source is not open season on American IP," said Bessent.
"When PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table."