US Treasury Threatens Sanctions Over Moonshot's “Distillation” of Fable

US Treasury Secretary Scott Bessent has once again said that sanctions against Chinese artificial-intelligence developers “remain on the table.” The trigger was an accusation from the White House: officials claim that Moonshot illegally “distilled” Fable, the model developed by Anthropic, in order to build its own Kimi K3.

Distillation is a common training technique in which a smaller model learns from the outputs of a larger one. The method is considered a legitimate way to optimize models, but under certain conditions it can infringe intellectual-property rights. That is precisely what Moonshot stands accused of.

“Open source is no license to hunt American intellectual property,” Bessent wrote on X. According to him, if Chinese firms are carrying out “covert, industrial-scale distillation attacks” that cross the line into IP theft, sanctions and an Entity List designation will come into play. Earlier in the week he had promised that the authorities would examine Chinese open models for signs of such theft.

Export controls and Nvidia servers

Bessent's statement came shortly after remarks by Michael Kratsios, head of White House science and technology policy. He accused Moonshot of large-scale distillation of American models and added that the company had acquired Nvidia servers with GB300 chips, as well as gained access to such systems in Thailand — likely to train its models. The GB300 accelerators belong to the Blackwell generation, whose sale to Chinese companies is prohibited, raising the question of a violation of export restrictions.

Experts' doubts

Some specialists are skeptical of the very premise. Fable became publicly available only on July 1, and Moonshot released Kimi K3 with open weights just last week — too short a span to explain the advanced capabilities of the new model through distillation from Fable alone.

At the same time, Kimi K3's success strikes at the business models of leading American labs: it casts doubt on the justification for the enormous capital investments underpinning the race for frontier AI. The episode has also sharpened a broader debate in Washington about the influx of Chinese open models. For instance, Dean Ball, a former White House AI adviser and now head of strategy at OpenAI, argues that the United States should restrict or effectively ban the use of Chinese open-weight models in order to preserve its technological edge and reduce national-security risks.

TechCrunch has reached out to Moonshot and the US Treasury for comment.

Source: TechCrunch