AI Industry Urges US to Avoid Broad Open-Weight Model Curbs

Chinese AI labs’ increasingly capable and cheaper open-weight models are challenging the economics of U.S. providers that keep their frontier systems closed. Washington is weighing how to respond after the White House accused Moonshot AI of using model distillation to copy Anthropic’s Fable in training Kimi K3. The debate matters because distillation — using one model’s outputs to improve another — is also a common development tool, raising the risk that an intellectual-property crackdown could spill into broad controls on downloadable models.

On July 24, 25 companies including Hugging Face, Meta, Microsoft, Mistral and Nvidia signed an open letter urging policymakers to avoid premature restrictions on open-weight AI. The group said unlawful extraction from closed systems should be handled through targeted legal and commercial frameworks, not sweeping limits on distillation or open models. Replit CEO Amjad Masad said banning Chinese open models would amount to banning open models generally, citing an ecosystem in which Thinking Machines Lab’s Inkling was trained with help from Moonshot’s Kimi 2.5.