Open AI Model Letter: Nvidia, Microsoft, Meta Sign On

Nvidia CEO Jensen Huang finally joined X — and used his first-ever post to wade into the sharpest policy fight in artificial intelligence. On July 24, 2026, Huang shared an open letter, signed by 25 organizations, arguing that the United States should protect and promote open-weight AI models rather than restrict them. “AI will transform every industry, power every company, and be built by every country,” Huang wrote. “Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models.”

The timing was pointed. Washington is actively debating whether to curb access to powerful Chinese open-weight systems, and the letter reads as a direct intervention in that debate from the companies with the most to gain from an open ecosystem — starting with the one that sells the chips everyone trains on.

Who signed — and who didn’t

The coalition spans chipmakers, cloud providers, open-source foundations, and venture firms. Signatories include Nvidia, Microsoft, Meta, IBM, Dell Technologies, Palantir, Mistral, Mozilla, Hugging Face, Perplexity, Replit, the Linux Foundation, Andreessen Horowitz (a16z), and Y Combinator, among others totaling 25 organizations.

The absences are as telling as the signatures. OpenAI, Anthropic, and Google — the three largest developers of closed, frontier-scale models — did not sign. That split maps almost perfectly onto commercial interest: the companies whose businesses depend on proprietary model access stayed out, while those that sell infrastructure, distribute open weights, or build on top of them lined up behind the letter. (Reports indicate OpenAI separately endorsed related language warning against “premature restrictions,” but it is not among the letter’s signatories.)

The argument

The letter’s core claim is that open weights are a strategic asset for American competitiveness, not a liability. It makes several connected arguments:

• Access. Open-weight models let startups, universities, hospitals, researchers, and small businesses build capable AI without the budgets required to rent frontier-scale proprietary systems for every workload. Openness, the signatories argue, widens who gets to participate in the AI economy.
• Cost. Organizations can run smaller, specialized models for common tasks instead of paying frontier prices for everything — a practical efficiency argument that dovetails with the rise of the small language model.
• Security and sovereignty. Counterintuitively, the letter contends that open weights strengthen cybersecurity by letting defenders inspect and harden models, and enable national “sovereignty” by letting countries build on systems they can host and control themselves.
• Innovation and resilience. Concentrating AI development in a few closed systems, the argument goes, makes the whole ecosystem more fragile; a mix of frontier closed and frontier open models maximizes competition and diffusion.

Huang reached for history to make the point, warning Washington not to repeat a mistake the software industry narrowly avoided in the 1980s — when heavy restriction of foundational technology could have strangled the open software movement that later powered the internet. The framing casts open-weight AI as the modern equivalent of open standards: infrastructure too important to lock down.

The China subtext

The letter cannot be read apart from the geopolitics driving it. Washington has been weighing whether to restrict access to Chinese open-weight models after Moonshot AI’s Kimi K3 — released July 16 and ranked among the most capable models anywhere — put a Chinese lab at the open-weight frontier. That release arrived alongside White House accusations that Chinese developers had improperly distilled knowledge from American models, sharpening the policy question: does restricting open weights protect U.S. advantage, or simply cede the open ecosystem to Beijing?

The signatories’ answer is that clamping down would backfire — pushing developers toward foreign open models while kneecapping the domestic open-source community that keeps American AI competitive. Their critics counter that freely available frontier weights lower the barrier for misuse and are effectively impossible to claw back once published, a tension that runs through the ongoing federal review of frontier AI policy.

What it means

The letter is less a technical document than a lobbying salvo, and reading it that way explains almost everything about who signed.

The alignment is commercial. Nvidia sells accelerators regardless of whether the models trained on them are open or closed — but a thriving open ecosystem multiplies the number of organizations training and serving models, which means more chips sold. Meta and Mistral publish open weights as a core strategy; Hugging Face, Mozilla, and the Linux Foundation are the open ecosystem. For them, “protect open models” is “protect our business.” The closed-lab holdouts — OpenAI, Anthropic, Google — have the opposite incentive, and their silence is a position.

The policy stakes are real. With open-weight models steadily closing the capability gap on their closed rivals, the question of whether the U.S. government treats open weights as an asset or a threat will shape where the next generation of AI development happens. A restrictive posture would hit domestic open-source labs first; a permissive one accepts that frontier-grade weights, once released, spread everywhere — including to adversaries.

What to watch. Whether the White House and Congress move toward export-style controls on frontier open weights, or toward the “keep it open” framing this coalition is pushing. Watch, too, whether any closed lab breaks ranks to sign a future version — and whether the China distillation fight hardens into concrete rules. Huang’s decision to spend his debut post on this, rather than a product, signals how central the fight has become: for Nvidia, the shape of AI regulation is now a first-order business risk, worth breaking a lifetime of social-media silence to influence.