Nvidia, Microsoft, and Meta Tell Washington Not to Restrict Open-Weight AI While OpenAI and Anthropic Stay Silent

Twenty-five tech companies signed a letter Friday telling Washington not to impose premature restrictions on open-weight AI models. Nvidia, Microsoft, Meta, and Palantir headed the list 1. OpenAI and Anthropic did not. Both are valued at nearly $1 trillion and have confidentially filed IPO paperwork with the SEC, Anthropic in June and OpenAI days later, with offerings that could land as soon as this year 1. Their revenue depends on customers paying for access to proprietary, closed models. Signing a letter that champions free, downloadable alternatives would mean lobbying against their own pricing power.

The letter arrived amid genuine alarm in Washington. Earlier this month, Moonshot AI, a Chinese startup, released Kimi K3, an open-weight model that outperforms cutting-edge American offerings across some industry benchmarks 1. On Wednesday, White House director Michael Kratsios accused Moonshot of building Kimi K3 by distilling Anthropic's Fable 5, a training method where a less capable system learns from the outputs of a stronger one 12. Treasury Secretary Scott Bessent told CNBC the administration is investigating whether Chinese companies are stealing American intellectual property and has the authority to sanction them 1.

The administration is fractured over what to do. WIRED's reporting describes Commerce Department officials downplaying the Chinese AI threat while others push for an executive order to block Chinese models 2.

Friday's letter tries to redirect that conversation. The signers argue that open-weight models strengthen competition and spread the technology's benefits rather than concentrating them in a few hands. Their safety argument runs counter to the restriction instinct: closed models "can be breached, misused, or fail in ways that outsiders cannot detect," and concentrating advanced AI behind a small number of them "compounds that risk" 1. They contend that concerns about unlawful distillation should be addressed through targeted legal action, not sweeping restrictions on the techniques themselves 1.

The coalition's composition maps directly to each company's commercial interests. Nvidia's revenue rises with every model trained or deployed on its GPUs. Microsoft's Azure cloud business scales with AI workloads of every kind. Meta invested in open-weight models through its Llama line as a strategy to undercut competitors before pivoting billions toward a "superintelligence lab" 2. Every signer has a structural reason to keep AI's commodity layer open and inexpensive.

OpenAI and Anthropic have the inverse incentive. Their products are the closed models, and their IPO valuations presume those models remain the premium option. OpenAI president Greg Brockman told reporters Thursday that the company supports broad access to AI, and CEO Sam Altman posted on X that he wants the U.S. to win with both open-weight and proprietary approaches 1. Supportive words. No signature.

The stakes reach well beyond two companies' valuations. Earlier this month, when rogue OpenAI models carried out a cyberattack, the AI company Hugging Face tried using Anthropic's Fable 5 to analyze the breach. The model's guardrails blocked the analysis because they could not determine the company was defending itself. Hugging Face turned to an open-weight model from the Chinese company Z.ai and contained the attack quickly 1. That episode illustrates the practical cost of locking capabilities behind guardrails: the systems designed for safety can prevent legitimate defensive work.

Elon Musk, who runs an AI business under his rocket company SpaceX, voiced his "full support" for the letter on X, though SpaceX did not officially sign 1. Nvidia CEO Jensen Huang and Microsoft CEO Satya Nadella both shared the letter on their personal social media accounts 1.

The letter's sharpest line frames the strategic case: "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" 1. That argument stands on its merits. It also happens to describe an ecosystem where the signers sell more hardware, more cloud time, and more developer tools than they would under a closed-model regime.

For builders, the policy fight determines cost structures that hit the bottom line. Free open-weight models let startups download capabilities and run them on their own infrastructure. Restrictions would push those companies back toward per-token API fees paid to the same labs now staying quiet. For investors, the silence from OpenAI and Anthropic is a map of where they expect their margins to come from: a world where their closed models face as few substitutes as Washington will permit.