OpenAI, Microsoft, Nvidia and friends tell Washington: don’t crack down on open-weight AI

On Friday, over two dozen major tech companies, organizations, and platforms released an open letter defending open-weight AI models, and it lands squarely in Washington's inbox. The list includes Nvidia, Microsoft, Meta, OpenAI, Mistral, and others, totaling 32 signatories as of Friday evening, according to the published roster.

This is not abstract. In his first-ever post on X, Nvidia CEO Jensen Huang wrote that “AI will transform every industry, power every company, and be built by every country,” and argued that open models “strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.” Huang’s message is basically the thesis of the letter: open-weight systems should be treated as a strategic infrastructure, not something the government tries to throttle.

Why the sudden lobbying moment? The letter comes at a time when top Trump administration officials have suggested they might take action against Moonshot AI, the creator of the Kimi K3 described as “the world's most powerful open-weight AI model.” The framing in the article is that the US could move against China-based open model makers like Moonshot AI either directly or by significantly impeding them.

That backdrop matters for decision-makers because it pulls open-weight AI into the same policy fight as export controls, cybersecurity, and IP enforcement. The signatories essentially argue that policymakers should not aim for broad crackdowns on open-weight model development. They contend that the US should build “a strong, open ecosystem that diffuses into every sector,” writing that their AI leadership “will be judged” not by a single frontier model, but by whether open ecosystems spread across industries.

Regulation risk is also being shaped by how the industry talks about “distillation.” Distillation is described in the article as training a less powerful model on the outputs of a more powerful one, a technique that is controversial but common in AI development. The immediate trigger for attention is a claim from the White House. On Wednesday, Michael Kratsios, director of the White House Office of Science and Technology Policy, publicly accused Moonshot of developing Kimi K3 by distilling the model from Anthropic’s Fable 5.

The letter does not ignore the controversy around distillation. It defends distillation more broadly, but the article notes it also “opens the door” to targeted crackdowns on companies that are seen as bad actors. The signatories write, “Unlawful efforts to extract value from closed models raise legitimate concerns,” and argue 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.” In other words: protect open model development, but aim enforcement at unlawful behavior, not at the general technique.

You can also see why this coalition is unusually heavy. The article points out that when the letter was initially published, the biggest absences were OpenAI and Anthropic, seen as leaders in the frontier model race. By Friday evening, OpenAI appeared to have signed on, as it was added to Microsoft’s published list of signatories. The article adds that OpenAI did not respond to a request for comment about signing, and it also details public friction: a community note under Sam Altman’s post claimed OpenAI had “refused” to sign, while Altman wrote on X earlier on Friday that he was “glad to see” support for open weight models and that he “want[s] the US to win in AI both in open source and proprietary models.”

Anthropic is another missing name. Like Google, the article says Anthropic did not sign the letter, and it references the ongoing competitive tension by contrasting models and approaches. It notes that unlike Anthropic’s Fable 5 and OpenAI’s GPT-5.6-Sol, Kimi K3 is an open-source model that “will soon be available for download and customization.” The strategic implication is straightforward: open-weight models can spread quickly, and if policymakers treat them as a national security threat, the market could freeze or fragment. If policymakers treat open-weight models as a strategic advantage, innovation diffuses faster across more players.

The coalition also tries to connect open-weight AI to the history of open software. The letter draws an analogy to how open software changed computing in the 1980s, arguing that open-source did more than lower costs, it created “a shared foundation of knowledge” that generations of engineers and entrepreneurs used to build “institutional sovereignty.” The article adds that while China has dominated development of open-weight models, the US has an “open-source” legacy. In this telling, the letter is both a defense and a dare: the US needs to lead in diffusion, not only in frontier labs.

The roster itself reads like a who’s-who of AI infrastructure and enterprise tech, spanning model and platform builders to security and developer tooling. It includes the American Innovators Network, Andreessen Horowitz, GitHub, Hugging Face, IBM, the Linux Foundation, and Mozilla, along with enterprise and security names like Box, Cisco, CrowdStrike, Dell Technologies, Palo Alto Networks, and ServiceNow. It also features companies tied to AI deployment and marketplaces like DoorDash and Y Combinator, plus specialized AI builders like Cohere, Arcee AI, Arena, Black Forest Labs, Fireworks AI, and Perplexity. The specific mix matters: this is not just researchers arguing about code. It is companies that would live with the compliance burden if Washington cracks down.

So what does this mean for the people who need to decide this week: CEOs, policy leads, and board members managing risk? The letter is asking for a particular kind of regulatory design. It is explicitly advocating against “sweeping restrictions” and in favor of “targeted legal and commercial frameworks.” If the US does move, the signatories are signaling that open-weight AI is not a niche product category, it is an ecosystem, and they want the rules to preserve the ecosystem while still addressing alleged misuse. For any company watching the regulatory weather, the message is clear: the next phase of AI policy could be about which models, which techniques, and which supply chain members get to operate, not just about who has the best benchmark.