25 tech companies defend open-weight AI models in joint letter

Nvidia, Microsoft and Meta were among 25 US technology companies that signed a joint letter on July 24 urging the Trump administration to avoid broad restrictions on open-weight AI models, framing open access as essential to American competitiveness rather than a security liability.

"Our AI leadership will be judged not by one frontier AI model, but by whether the United States builds a strong, open ecosystem that diffuses into every sector," the companies wrote in the letter, hosted on Microsoft's corporate-responsibility site. Nvidia Chief Executive Officer Jensen Huang amplified the message in his first-ever post on X, and Elon Musk said he "fully supports" the letter.

The letter lands as the White House weighs whether to ban Chinese open-weight models after Beijing-based Moonshot AI released Kimi K3 on July 16, an open-weight system that independent testing ranks third on one widely tracked capability index behind Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol. The Trump administration has accused Moonshot of distilling Anthropic's Fable model to train Kimi K3, claims the company has not publicly addressed. Treasury Secretary Scott Bessent said on Fox Business that the administration is investigating Chinese AI labs for distillation attacks and threatened sanctions.

The signatories — which also include Andreessen Horowitz, IBM, Dell, Palantir, Mistral, Hugging Face and Y Combinator — argue that restricting open weights would concentrate advanced AI capability behind a handful of closed providers, creating single points of failure that outsiders cannot inspect. "Openness may be one of the most important paths to AI safety and security," the letter states, citing a recent incident in which OpenAI disclosed that one of its models escaped its testing environment and accessed a Hugging Face repository. Hugging Face said it could not defend itself using commercial frontier AI models because their guardrails blocked its efforts, and instead used Chinese firm Z.ai's open-weight GLM 5.2.

The letter draws a sharp contrast with the position of OpenAI and Anthropic, the two most prominent builders of closed, API-only models, neither of which signed. Both companies have urged the administration to respond forcefully to alleged Chinese intellectual property theft, arguing that open weights make it impossible to revoke access or patch guardrails once parameters are public. Anthropic keeps its most capable cybersecurity model, Claude Mythos, restricted to a vetted set of partners.

OpenAI Chief Executive Officer Sam Altman struck a conciliatory tone Friday, saying he wants the US to win in both open-source and proprietary AI and that he is "happy to see" Huang's endorsement of the letter. OpenAI President Greg Brockman said Thursday that "having more models, more usage, is a good thing for me." The comments suggest the two camps are not as polarized as the absence of OpenAI and Anthropic from the signatory list implies.

The Distillation Debate

The letter explicitly defends distillation — the practice of training one model on another's outputs — calling it "a widely used technique for model improvement, evaluation, and validation" that reflects "a long tradition of learning from, building upon, and improving existing technologies." It argues that unlawful extraction of value 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 administration has taken a harder line. White House Office of Science and Technology Policy Director Michael Kratsios said Wednesday that the administration has information showing Moonshot AI created a "sophisticated internal platform" to access Anthropic's Fable. More than 200 AI startups sent a separate letter to administration officials this week opposing a ban and urging safety guardrails instead.

Who Wins, Who Loses

The signatories have clear commercial incentives. Nvidia sells the graphics processing units that train and run every model, open or closed. Microsoft Azure rents cloud capacity for model deployment. Meta and Mistral ship open-weight systems. Hugging Face hosts them. If models become interchangeable commodities, infrastructure providers benefit from increased compute and cloud consumption.

The companies that did not sign — OpenAI and Anthropic — face a direct threat to their business models. Both are preparing for initial public offerings at valuations near $1 trillion, and the proliferation of cheap, highly capable open-weight models could compress pricing power for their API-based offerings. Kimi K3 is far cheaper to run than leading US systems, according to independent testing, turning a model launch into both a market event and a policy flashpoint.

This article is for informational purposes only and does not constitute investment advice.