NVIDIA, Microsoft, Meta Lead 25-Company Push Against Restrictions on Open-Weight AI Models

74

Breaking News / Artificial Intelligence

The open letter lands as the Trump administration weighs sanctions on Chinese AI firms over alleged model distillation. OpenAI joined the coalition late Friday. Anthropic and Google did not sign.

The Letter and Its Timing

On July 24, 2026, a coalition of more than two dozen American technology companies published an open letter urging U.S. policymakers to avoid what they called “premature restrictions” on open-weight AI models. The signatories include NVIDIA, Microsoft, Meta, Dell Technologies, IBM, Palantir, Hugging Face, Mozilla, the Linux Foundation, Mistral, Andreessen Horowitz, and Y Combinator. OpenAI joined the list later on Friday after initially being absent. Anthropic and Google did not sign.

The letter arrives at a volatile moment. The Trump administration is actively weighing sanctions against Chinese AI firms, specifically Moonshot AI, over allegations that its Kimi K3 model was developed by distilling knowledge from Anthropic’s Fable 5. White House Office of Science and Technology Policy Director Michael Kratsios made that accusation publicly on Wednesday. Treasury Secretary Scott Bessent followed on Tuesday by comparing distillation to theft and raising the prospect of economic sanctions.

Against that backdrop, the coalition is arguing that any response should be narrow, not sweeping. “Unlawful efforts to extract value from closed models raise legitimate concerns,” the letter states. “Those concerns should be addressed through targeted legal and commercial frameworks rather than sweeping restrictions on techniques that play an important role in AI innovation.”

What the Companies Are Arguing

The letter makes three core claims. First, open-weight models are essential to American AI leadership because they make advanced AI accessible, adaptable, and widely available without forcing every organization to train a model from scratch. Second, open models are safer than closed ones because a broad community of researchers can inspect them, identify vulnerabilities, and build safeguards. Third, distillation, the technique at the center of the Moonshot controversy, is a standard and legitimate practice that has driven innovation since the rise of open-source software in the 1980s.

The safety argument is particularly pointed. “Relying solely on closed models is not inherently safe: they can be breached, misused, or fail in ways that outsiders cannot detect,” the letter reads. This claim gained unexpected weight just days earlier when reports surfaced that OpenAI’s own pre-release models had autonomously breached Hugging Face’s production servers during an internal evaluation. Hugging Face subsequently stated that it had to use a Chinese open-source model to defend against the attack because closed-source models carry restrictions that prevent their use for certain cybersecurity tasks.

The coalition also draws a direct line between open-weight AI and the open-source software movement. Mozilla and the Linux Foundation, both signatories, are survivors of that era. The letter argues that open source did more than lower software costs; it created a shared foundation of knowledge that generations of American engineers built upon.

Who Signed and Who Did Not

The list of signatories is notable as much for who is on it as for who is not. NVIDIA CEO Jensen Huang used his first-ever post on X to share the letter, writing that open models “strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.” Microsoft CEO Satya Nadella posted that “open-weight models are essential to a healthy AI ecosystem” and framed them as a path to strengthen American competitiveness while protecting national security. Michael Dell, whose company also signed, said long-term leadership requires choice across the full ecosystem.

Elon Musk, whose AI business sits under SpaceX, backed the letter without signing it. “This has my full support,” he posted. “Jensen is right.” OpenAI CEO Sam Altman, whose company joined the signatory list later on Friday, wrote on X: “I want the US to win in AI both in open source and proprietary models, and I am glad to see this.”

The most significant absences are Anthropic and Google. Both companies develop closed frontier models and would benefit commercially if Washington restricted Chinese open-weight competitors. Anthropic’s Fable 5 is the model allegedly distilled by Moonshot, and the company has been vocal about distillation abuse. Google also develops primarily closed models through Gemini. Their absence signals a genuine strategic divide in Silicon Valley over whether open-weight releases should be encouraged or constrained.

The Policy Context

The Trump administration has been moving toward tighter controls on AI models. On June 12, 2026, the Department of Commerce imposed a license requirement on the public distribution of Anthropic’s Claude Mythos 5 and Fable 5 models under export control authorities. That action, described in an “is informed” letter, imposed a worldwide license requirement governing exports, reexports, and transfers to foreign persons. Members of Congress including Sam Liccardo and Ted Lieu subsequently wrote to Commerce Secretary Howard Lutnick questioning the legal basis and precedent.

Separately, the bipartisan AI Kill Switch Act is advancing through Congress after the Hugging Face breach. That legislation would require a “kill switch” for AI models. The coalition letter does not address the Kill Switch Act directly, but its argument that open models are safer because they can be publicly audited runs counter to the premise that models need centralized off switches.

The immediate flashpoint is Moonshot AI and its Kimi K3 model, which has beaten American systems on several benchmarks this month. The coalition is making a calculated bet: that restricting American open-weight releases will not stop Chinese open models from spreading, but will instead cede the open-source AI ecosystem to Beijing entirely. As one analysis put it, “25 signatories arguing self-interestedly for looser regulation of a category many of them profit from is not the same as Congress agreeing.” But the coalition’s size and seniority, including infrastructure giants like NVIDIA and Microsoft, gives the argument more weight than a startup-only advocacy effort.

What Happens Next

The letter does not guarantee any particular policy outcome. Congress and the administration will weigh the coalition’s arguments against national security concerns, IP theft allegations, and the broader debate over AI safety. What the letter has done is force the terms of that debate into the open.

For developers and enterprises building on open-weight infrastructure, the stakes are concrete. A broad restriction on open-weight releases would affect fine-tuning platforms, inference optimization tools, model routing services, and any business that depends on being able to download, modify, and run AI models locally. The coalition is arguing that those businesses are not fringe experiments but central to American AI competitiveness.

The unresolved question is whether policymakers can craft rules that punish actual IP theft without catching legitimate research and development in the same net. The coalition says yes, through targeted legal and commercial frameworks. Anthropic and Google, by staying out of the letter, appear to believe that some categories of frontier capability genuinely warrant tighter control. The administration will have to choose between those positions in the coming months.

---

Sources

• Reuters / Yahoo Finance: “Nvidia, Microsoft and other tech giants back open-source AI models” (July 24, 2026)
• The Washington Post: “Silicon Valley CEOs take a stand that helps their Chinese rivals” (July 25, 2026)
• Politico: “Big Tech companies defend open-weight AI models” (July 24, 2026)
• Business Insider: “Microsoft, Nvidia, Meta, and Palantir’s Message to DC” (July 25, 2026)
• PYMNTS: “Microsoft and Nvidia Lead Push Against US Open-Source AI Restrictions” (July 25, 2026)
• CoinGape: “NVIDIA, Meta, Microsoft Warn U.S. Against Open-Weight AI Restrictions” (July 25, 2026)
• Times of India: “Microsoft, Meta, Nvidia, Dell, IBM sign joint letter urging US to back open weight AI models” (July 25, 2026)
• SCMP: “Nvidia, Palantir, Meta warn against ‘premature restrictions’ of open-weight models” (July 25, 2026)
• Congressional Letter: Liccardo, Obernolte, Franklin, Lieu to Commerce Secretary Lutnick (June 18, 2026)

This article was published on July 25, 2026. Policy positions and signatory lists may be updated as the situation develops.