Treasury threatens sanctions, claims Moonshot distilled Anthropic's Fable (techcrunch.com)

U.S. Treasury Secretary Scott Bessent has issued a stern warning to Chinese AI companies, particularly targeting Moonshot, which has been accused of improperly distilling Anthropic’s Fable model. The contentious process of model distillation involves a smaller AI model learning from a larger one, a technique that can tread the line of intellectual property infringement. Bessent emphasized that open-source frameworks cannot be exploited to undermine American intellectual property, and indicated that sanctions could follow if violations of U.S. export-control rules regarding technology, such as Nvidia's restricted servers, are confirmed. The allegations against Moonshot have raised significant concerns within the AI/ML community, as the firm's recent release of the Kimi K3 model challenges the competitiveness of U.S. AI labs. Experts are divided on the legitimacy of claims that K3 was primarily developed through distillation of Fable, especially given Fable's recent public availability. This situation not only amplifies scrutiny over the use of Chinese AI models but also reflects a broader debate on national security and technological advantage in the rapidly evolving AI landscape. As the U.S. government considers regulatory actions, the implications for innovation, collaboration, and competition within the global AI market could be profound.