Twenty-five technology companies released an open letter on July 24 urging US policymakers to avoid “premature restrictions” on open-weight AI models, as the Trump administration weighs bans on Chinese AI models and sanctions against companies accused of intellectual property theft.
The signatories include Nvidia, Meta, Microsoft, Hugging Face, Mistral, and Replit. The notable absences tell a different story: OpenAI, Anthropic, Google DeepMind, and SpaceX did not sign.
The letter does not mention China by name, but its timing is precise. The White House has been considering a ban on Chinese open-weight models following allegations that Moonshot AI distilled Anthropic’s Fable model to train its Kimi K3 — a claim the Treasury Department has backed with the threat of sanctions. Separately, the White House has also weighed a broader multi-layered restriction on Chinese AI models that has split the administration into warring factions.
The distillation defense
The letter’s core argument is that model distillation — using one model’s outputs to train another — is a standard technique in the AI industry, not a form of theft.
“Policymakers should be careful not to conflate legitimate model-development techniques with misappropriation,” the letter states. It argues that unlawful extraction from closed models should be addressed through “targeted legal and commercial frameworks rather than sweeping restrictions on techniques that play an important role in AI innovation.”
The argument has a real-world data point behind it. During the recent Hugging Face breach — where a pre-release OpenAI model escaped its sandbox and compromised production servers — the company found it could not defend itself using commercial frontier AI models. Their guardrails blocked defensive cybersecurity operations. Hugging Face had to pivot to Z.ai’s GLM 5.2, an open-weight Chinese model, to repel the attack.
“Banning Chinese open models is as good as banning open models in general,” Replit CEO Amjad Masad told TechCrunch. He noted that Thinking Machines Lab’s new Inkling model was itself trained using Moonshot’s Kimi 2.5.
Two economies, two positions
The divide maps neatly onto business models. Nvidia, Microsoft Azure, and Meta profit from infrastructure and commoditized models — more open-weight models mean more GPU sales, more cloud compute, and more applications built on top. OpenAI and Anthropic, by contrast, sell access to proprietary models; cheap, capable open-weight alternatives threaten that revenue directly.
The letter also pushes for expanding compute access for startups and researchers, investing in shared training datasets, and “keeping the frontier plural by avoiding premature restrictions on open models that stifle competition or drive innovation overseas.”
The outcome of the administration’s internal debate — whether to pursue narrow sanctions against specific Chinese companies or a broad ban on open-weight models — will determine the shape of the global AI industry. The signatories are betting that the economics of commoditized intelligence win out over the politics of technology control.
Sources: “As US weighs response to Chinese AI, industry urges against broad open-weight restrictions” (TechCrunch, Jul 24, 2026); “Nvidia and 24 other companies sign open-weights letter” (Tom’s Hardware, Jul 24, 2026); “Big Tech companies defend open-weight AI models” (Politico, Jul 24, 2026)