This was the first time the U.S. government itself — rather than a private AI lab — has directly named a specific Chinese company as the perpetrator of industrial-scale model extraction .
Treasury Secretary Scott Bessent had already previewed the administration's stance on July 21, saying the U.S. would carefully examine Chinese open-source AI models for IP theft and take action . On July 22, hours after Kratsios's announcement, Bessent doubled down, posting on X: "Open source is not open season on American IP. 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" . He confirmed sanctions and export restrictions remain "on the table" for Moonshot and other Chinese AI firms .
On July 27, China's Ministry of Commerce issued a formal response, accusing the United States of "AI hegemonism" and applying double standards . The ministry stated the U.S. threats lack "factual and legal grounds," calling them a "typical example of AI hegemony" . It warned that China will take "all necessary measures" — meaning countermeasures — to protect its companies if the U.S. imposes sanctions or trade restrictions .
The US claim is publicly disputed by independent AI experts. TechCrunch reported on July 23 that multiple experts said "exploiting Anthropic's Fable isn't how Kimi K3 got so good," casting doubt on whether distillation of Fable 5 could plausibly produce Kimi K3's performance profile . The U.S. government has not released technical evidence supporting the accusation . Moonshot has not publicly admitted to the practice.
The Moonshot case sits within a broader, sometimes contradictory Trump administration posture on China AI:
The current escalation — using model distillation as a trigger for Entity List designations — represents a significant new enforcement vector beyond the existing chip-focused export controls.