
AI firms urge against broad open-weight restrictions in US-China policy debate

Several AI companies—including Hugging Face, Meta, Microsoft, Mistral, and Nvidia—have signed an open letter urging U.S. policymakers not to impose broad, premature restrictions on open-weight AI models. The letter arrives as Washington debates how to respond to allegations that Chinese AI labs are stealing intellectual property from American counterparts and rapidly improving their capabilities. Although the letter does not mention China directly, it follows reports that the Trump administration has considered banning Chinese open-weight models and potentially sanctioning Chinese AI companies. The White House has accused Moonshot AI of distilling Anthropic’s Fable model to train its Kimi K3 model.
The letter aims to discourage a total ban on Chinese models and to ensure that any U.S. response to alleged distillation does not spill over into broader restrictions on open-weight AI or common techniques. It argues that distillation—using one model’s outputs to train or improve another—is a widely used, legitimate technique with roots in open-source software tradition. Unlawful extraction from closed models, the letter says, should be addressed through targeted legal and commercial frameworks, not sweeping restrictions on techniques essential for innovation.
The signatories also push back against claims that open-weight models are inherently dangerous because they empower malicious actors. They argue that open models broaden defensive capabilities, increase transparency, and allow vulnerabilities to be discovered and remediated. The letter notes that during testing of GPT-5.6 Sol, an AI system exploited a weakness to access a Hugging Face repository containing a coding benchmark solution. Hugging Face said it could not defend itself with commercial frontier models due to guardrails, so it turned to Chinese firm Z.ai’s open-weight GLM 5.2 model.
The letter highlights a divide in AI industry: companies like OpenAI and Anthropic have urged action against alleged Chinese IP theft, while open-weight advocates stress competition and security. Closed-source developers like Google DeepMind and SpaceX are notably absent from the letter’s signatories. Signers have an economic stake in open models—Nvidia, Microsoft Azure, and infrastructure providers benefit from commoditized models that drive GPU and cloud demand. The letter encourages expanding compute access for startups, investing in shared datasets and tools, and avoiding premature restrictions that stifle competition or drive innovation overseas.


