The signatories turn the safety case back on the closed-model leaders, warning that routing AI through a few proprietary systems creates single points of failure — even as Washington reportedly weighs barring Chinese open-weight models.
Twenty-five companies and organizations, among them Nvidia, Meta, Hugging Face and the Linux Foundation, published an open letter on Friday titled Open Weights and American AI Leadership, urging U.S. policymakers not to restrict open-weight AI models — systems whose fundamental parameters are released publicly, leaving anyone free to download, self-host and modify them.
Nvidia chief Jensen Huang fronted the effort with his first-ever post on X, writing that the world needs leading closed models and leading open models alike.
The most conspicuous absences are OpenAI, Anthropic and Google, the front-runners in closed frontier models sold through paid APIs, who would gain if restrictions removed their cheapest open rivals. The signatories have stakes of a different kind: Nvidia profits from wide adoption, Meta builds its own models while trailing the leaders, and Hugging Face lives on distributing open ones.
The letter, Open Weights and American AI Leadership, argues that relying solely on closed models carries no built-in safety, because such models can be compromised, misused or fail without outsiders noticing. Openness, it says, may rank among the most important routes to AI safety and security.
The appeal lands amid a live policy fight. The Trump administration has reportedly weighed prohibiting Chinese open-weight models after the July 16 release of Moonshot AI's Kimi K3, and this week the White House science and technology director, Michael Kratsios, accused Moonshot of building it by distilling Anthropic's Fable 5. The letter defends distillation as a legitimate technique, arguing that unlawful extraction deserves narrowly aimed legal and commercial measures rather than blanket bans.
OpenAI did not push back. Its chief executive, Sam Altman, wrote on X that he was pleased to see the backing for open weights, adding that he wanted the US to win in AI across both open-source and proprietary models.
Every signatory still has something to gain if open AI takes hold, The Next Web notes, and the letter does not resolve whether the U.S. can still win an open-model race it is currently losing to China.