Policy The AI Industry Just Split in Public Over Open Models On 24 July 2026, 25 companies signed a letter urging Washington not to restrict open-weight AI. Nvidia, Microsoft, and Meta signed. OpenAI, Anthropic, and Google did not. The split is the story. Last updated: 25 July 2026 The short version A letter titled "Open Weights and American AI Leadership," dated 24 July 2026, was signed by 25 companies and groups and hosted on Microsoft's website. Signers include Nvidia, Microsoft, Meta, Mistral, Hugging Face, Palantir, IBM, Dell, Mozilla, the Linux Foundation, and Y Combinator. OpenAI, Anthropic, Google, and Amazon did not sign. None of the signers sells access to a closed frontier model of its own. Nvidia CEO Jensen Huang joined X and used his first-ever post to share the letter. The letter's sharpest passage defends distillation as legitimate and asks Washington not to ban it. It never names China, Moonshot, or DeepSeek. Disclosure: this story concerns a policy dispute in which Anthropic is a named absent party and the subject of a direct comment by Nvidia's CEO. This newsroom uses Anthropic's Claude in production. What the letter says The letter argues that American AI leadership will be judged not by any single frontier model but by whether the country builds a broad, open ecosystem. Open-weight models, defined in the document as models anyone can download, inspect, modify, and run on their own infrastructure, are cast as the foundation for that ecosystem. The case rests on four claims: open weights widen access to the AI economy, they strengthen competition, they give customers protection against vendor lock-in, and, the letter's most contested argument, they may improve safety rather than weaken it. On that last point the signers meet the security objection head-on. They concede open weights carry real risks, since once released the weights are beyond the developer's control. But they argue closed models are not inherently safer, because they can be breached or misused in ways outsiders cannot detect, and that concentrating capability behind a few closed systems creates single points of failure. The framing reaches back to the 1980s open-source software movement, which the letter credits with building the foundation that now runs most of the internet and much of the US government's own systems. Who signed, and who did not The signal is in the roster. The 25 signers span chipmakers, cloud and server vendors, open-model labs, security firms, and venture funds. What unites them is that none sells access to a closed frontier model as its core product. The companies that do, OpenAI, Anthropic, and Google, are absent, and so is Amazon. Signed the letter Did not sign Nvidia, Microsoft, Meta, Dell, IBM OpenAI Mistral, Hugging Face, Mozilla, The Linux Foundation Anthropic Palantir, CrowdStrike, Box, ServiceNow, Perplexity, Replit Google Andreessen Horowitz, Y Combinator, Emergence Capital Amazon The full signatory list also includes American Innovators Network, Arcee AI, Arena, Black Forest Labs, Mariana Minerals, Reflection, and Telnyx. The pattern holds down the list: every signer benefits from a plural, open ecosystem it can build on or sell into, and the absent four each hold a closed model or a business that a flood of free open weights would pressure. The letter is an argument, and it is also an interested one. Its value is that it puts the open camp's position on the record in one place, at the moment Washington is deciding. The distillation clause is the real policy fight The letter's most consequential paragraph is about distillation, the practice of training one model on another model's outputs. The signers defend it as a widely used, legitimate technique for improving and evaluating models, part of a long tradition of building on existing technology. They draw a line between that and what they call unlawful efforts to extract value from closed models, and they argue those should be handled through targeted legal and commercial frameworks rather than sweeping restrictions. That is a direct intervention in the live fight in Washington. The Trump administration has weighed limiting US access to Chinese open models, and White House adviser Michael Kratsios has accused Moonshot AI of distilling a US model to build its Kimi K3. claim Notably, the letter never mentions China, Moonshot, or DeepSeek by name. It argues the principle while leaving the specific target unspoken, which lets every signer defend distillation in general without endorsing any particular lab's conduct. Huang goes further than the letter does Huang put his own name to the cause in two ways. He joined X and used his first post to share the letter, writing that open models strengthen safety, cybersecurity, innovation, and sovereignty, and that "the world needs both frontier closed models and frontier open models." claim Two days earlier, in an interview with Axios in Fort Worth, he went past anything the letter says. Speaking about Anthropic's restricted cybersecurity model, he argued "Mythos should be available as a service," adding that "holding Anthropic back is not in the benefit of the United States." claim Worth noting: Nvidia is already one of the partners with access to Mythos, through Anthropic's Project Glasswing program, as this publication reported when the US cleared Mythos 5 for critical-infrastructure defenders in June. The letter itself never names Mythos or Anthropic; that was Huang, on his own, in a separate interview. Does this change anything yet? Not on its own. A letter is advocacy, not law, and no statute changed on 24 July. What it does is draw the industry's dividing line in public and hand policymakers a single document that states the open camp's case, timed to the moment Washington is weighing restrictions on Chinese open models. The absent signatures matter as much as the present ones: they mark which companies would rather the frontier stay closed, and which have bet their business on it opening. The next move belongs to the administration, and any hearing on open-model rules now has a 25-name counterweight on the record. Sources "Open Weights and American AI Leadership," 24 July 2026 (full letter text and 25-name signatory list), hosted by Microsoft. Axios, "Nvidia's Jensen Huang defends Chinese AI amid Kimi panic," 22 July 2026 (Huang's Mythos and open-model comments, on the record). Tom's Hardware, "Nvidia and 24 other companies sign open-weights letter," 24 July 2026 (signatory breakdown; letter does not name China, Moonshot, or DeepSeek). AI Race Facts, "US Clears Anthropic's Mythos 5 for Cyber Defenders," 27 June 2026 (Project Glasswing, Nvidia's partner access). Last updated: 25 July 2026. Figures and statements labeled claim are attributed to the named source and not independently audited. Positions in the letter are those of its signatories.