US threatens sanctions over AI model distillation by Moonshot AI
Washington is threatening sanctions and export restrictions against Chinese AI firm Moonshot AI, signaling a new front in the tech trade war that raises the stakes for American intellectual property protection.
The White House has accused Chinese artificial intelligence firm Moonshot AI of stealing proprietary technology from Anthropic to build its latest model, prompting threats of sanctions from the US Treasury.
Michael Kratsios, director of the White House Office of Science and Technology Policy, alleged in a post on X on Wednesday that Moonshot built an internal platform to covertly distill US models at scale. “Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem,” Kratsios said. “However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable.”
The accusation marks an escalation in how Washington polices the flow of AI capabilities, shifting focus from hardware export controls to the software and training methods underlying advanced models. For investors and tech executives, the prospect of sanctions introduces fresh geopolitical risk into the AI sector, potentially isolating Chinese firms from US-linked financial systems and supply chains.
Treasury Secretary Scott Bessent underscored this regulatory shift, warning that covert distillation attacks crossing into IP theft would carry severe consequences. “We support open-source AI and the innovation it unlocks. But open source is not open season on American IP,” Bessent said. “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.”
Despite the aggressive rhetoric from Washington, independent AI researchers have raised doubts about the technical plausibility of the specific claims against Moonshot. The White House asserts that Moonshot used Anthropic’s Fable 5 model to develop its Kimi K3 system, which launched on July 16.
Anthropic had only re-released Fable 5 on July 1 after briefly taking it offline to comply with US export controls. That leaves a narrow 15-day window for the alleged distillation to have occurred. “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.”
This skepticism was echoed by Dean Ball, OpenAI’s head of strategic futures, who said on Friday he did not believe the K3 model’s performance could be “explained away by distillation or anything like that.” Kimi K3 has emerged as one of China’s most capable AI models, intensifying Washington's anxiety over the pace of Chinese AI progress regardless of the technical debate.
