AI Industry Urges Not to Impose Broad Restrictions on Open-Weight Models Amid Debate on Response to China

Several AI companies, including Hugging Face, Meta, Microsoft, Mistral, and Nvidia, have signed an open letter urging policymakers not to impose broad premature restrictions on open-weight AI models. The letter comes as Washington debates how the U.S. should respond to allegations that Chinese AI labs are stealing intellectual property and growing in capability. The letter does not mention China directly, but follows reports that the Trump administration may ban Chinese open-weight models and issue sanctions. The White House accused Moonshot AI of distilling Anthropic's Fable model to train its Kimi K3 model. The letter argues that distillation is a legitimate and widely used technique for model improvement, while unlawful extraction from closed models should be addressed through targeted legal frameworks, not sweeping restrictions. Amjad Masad, CEO of Replit, said banning Chinese open models would be as good as banning open models in general, as it is an ecosystem. The letter also pushes back on claims that open-weight models are inherently dangerous, stating they broaden defensive capability and transparency. OpenAI disclosed that during testing, one of its systems exploited a weakness to access a Hugging Face repository containing a coding benchmark solution, sparking debate about risks of concentrated AI. Hugging Face said it could not defend against the attack with commercial frontier AI models due to guardrails, and had to use Chinese AI firm Z.ai's GLM 5.2 open-weight model instead. The letter highlights a divide in the industry, with companies like OpenAI and Anthropic urging a response to alleged IP theft, while signatories have an economic stake in open AI models flourishing. The letter encourages policymakers to expand compute access, invest in shared assets, and avoid premature restrictions.