- The Nvidia-hosted letter now lists 50 organizations, twice its launch total, after Google, OpenAI, AMD, Cisco and Cloudflare joined the coalition. [1][2] - The signatories want expanded access to computing resources and shared datasets while opposing broad restrictions on open models and legitimate uses of model distillation. [1] - The Trump administration has considered Entity List designations, procurement pressure and security-liability rules aimed at Chinese AI developers, but the reported proposals have not become an announced blanket ban. [3][4] - Anthropic remains a prominent nonsignatory and has argued that releasing powerful model weights creates safety risks because access cannot easily be revoked. Amazon and xAI are also absent, but absence alone does not establish opposition.
[1][5] An industry campaign against broad restrictions on downloadable AI models has doubled to 50 signatories, adding Google and OpenAI to a coalition that already included Nvidia, Microsoft, Meta, IBM and companies supplying the chips, cloud capacity and software used to run artificial intelligence.[1][2]
The expansion changes the picture from the letter’s launch on Friday, July 24. OpenAI and Google were absent from the original group of 25 but now appear in the Nvidia-hosted document. Anthropic, Amazon and Elon Musk’s xAI remain off the published list as of Sunday, July 26.[1][2]
The coalition spans the AI supply chain #
The current signatories are AI21, AMD, American Innovators Network, AMP, Andreessen Horowitz, Arcee AI, Arena, Baseten, Black Forest Labs, Block, Box, Cisco, Cloudflare, Cohere, CrowdStrike, Dell Technologies, DoorDash, Emergence Capital, Fireworks AI, Genspark, GitHub, Google, Hugging Face, IBM, Inferact, Interconnects AI, the Linux Foundation, Mariana Minerals, Meta, Microsoft, Mistral, Morph, Mozilla, Nebius, Nous Research, Nvidia, Ollama, OpenAI, OpenClaw, Palantir, Palo Alto Networks, Periodic Labs, Perplexity, Prime Intellect, Reflection, Replit, ServiceNow, Telnyx, Trajectory and Y Combinator.[1]
The roster includes model developers, chip designers, cybersecurity vendors, cloud and server suppliers, enterprise-software companies, venture firms and open-source organizations. Many have commercial interests in a market where customers can choose among models and run them on third-party clouds, rented GPUs or private infrastructure rather than buying access exclusively through a model developer’s service.[6]
The letter defines open-weight systems as models whose parameters can be downloaded, inspected, modified and run on privately controlled infrastructure. It acknowledges that released weights are difficult to recall or trace after modification, but argues that access also helps outside researchers test models, identify vulnerabilities and build cybersecurity defenses.[1]
The policy request targets broad restrictions and distillation rules #
The coalition asks policymakers to expand computing access for startups and researchers, invest in shared datasets and evaluation tools, and avoid what it calls premature restrictions that could reduce competition or push development overseas. It also argues that organizations should be able to retain control of their data and avoid dependence on a single model provider.[1]
A second request concerns distillation, in which developers use one model’s outputs to train, improve or evaluate another system. The signatories say legitimate distillation should remain available and that unlawful extraction from closed systems should instead be addressed through targeted legal and commercial enforcement.[1]
The document does not mention China, DeepSeek or Moonshot AI. Its timing, however, places it in a dispute over increasingly capable Chinese models such as Moonshot’s Kimi K3, DeepSeek’s systems and Z.ai’s GLM family. American developers and businesses are increasingly evaluating or using those models because they are downloadable and can be cheaper than leading proprietary services, according to Associated Press reporting.[7]
Washington is distinguishing openness from alleged theft #
Axios reported that administration officials had considered adding Chinese laboratories to the Commerce Department’s Entity List, issuing a federal security advisory, using procurement pressure and requiring American hosts to accept liability for breaches involving Chinese models. Earlier versions of those proposals were stopped by officials concerned about competition and innovation. The reporting described options under consideration rather than an announced blanket prohibition.[3]
The administration’s more recent public position supports open-weight development in principle while threatening sanctions or Entity List designations for what officials describe as covert, industrial-scale distillation of American technology. Officials have accused Moonshot of using Anthropic’s Fable model to develop Kimi K3. China has rejected accusations of illicit distillation as groundless.[4][7]
That distinction broadly overlaps with the coalition’s request for targeted enforcement, but substantial questions remain about where legitimate model development ends and misappropriation begins. Broadly written restrictions could also affect independently developed models, modified versions hosted by American companies or research access to downloadable weights.
Anthropic remains a prominent nonsignatory #
Anthropic has not signed and has not publicly provided a reason specific to this letter. CEO Dario Amodei has argued that open-weight models become harder to control once released because developers cannot readily revoke access, update safeguards or prevent misuse. Anthropic has supported stricter controls as models approach capabilities that could materially assist cyberattacks or biological-weapons development.[5]
OpenAI’s signature makes its position more nuanced. The company has warned Washington about powerful Chinese models and suspected distillation while also saying Chinese advances are not an argument against openness. Google, which publishes its Gemma open-weight family, also joined after initially being absent from the document.[1][2][5]
Amazon and xAI are not listed, but their absence should not be treated as a declared policy position. The immediate dispute is narrower than a vote for or against open AI: it concerns whether Washington can target alleged security threats and intellectual-property violations involving Chinese developers without broadly restricting technology used by American companies, startups and researchers.
Companies mentioned #
Further sources #
[[1] Open Weights and American AI Leadership, Nvidia-hosted coalition letter and cur… ↗](https://images.nvidia.com/pdf/Open-Weights-and-American-AI-Leadership.pdf)
[[2] Forbes, syndicated by Yahoo Tech: Nvidia Open Weights Letter Doubled To 50 With… ↗](https://tech.yahoo.com/ai/articles/nvidia-open-weights-letter-doubled-011942086.html)
[[3] Axios: The secret Trump administration battle to fight Chinese AI, July 20, 202… ↗](https://www.axios.com/2026/07/20/ai-us-china-open-source-kimi)
[[4] Axios: White House draws new AI line on China, July 24, 2026. ↗](https://www.axios.com/2026/07/24/white-house-ai-line-china)
[[5] Axios: OpenAI and Anthropic unite against China’s open models, July 22, 2026. ↗](https://www.axios.com/2026/07/22/openai-anthropic-open-models-trump-china)
[6] The Washington Post: Top tech firms urge U.S. government not to limit ‘open’ AI… ↗+1 more
The stories that matter, in one email. Free — unsubscribe anytime.