Twenty-five companies, including Nvidia, Microsoft, Meta, Andreessen Horowitz, Hugging Face, IBM and Dell, have told Washington not to restrict open-weight AI. Their letter, published Friday under the title "Open Weights and American AI Leadership," argues that limiting open models would not protect the United States and would instead hand the AI market to a small number of closed labs.
The case for open-weight models
Open-weight models are AI systems whose underlying weights, the billions of numbers that shape how a model responds, are published for free. Anyone can download them, run them on their own servers, and modify them.
That stands in contrast to closed models such as OpenAI’s GPT and Anthropic’s Claude, which are generally available only through a company’s paid app or API.
The letter compares the current moment in AI to the open-source software movement of the 1980s, saying the industry has reached a similar fork in the road. "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 letter says.
Huang and Nadella publicly back the letter
Nvidia CEO Jensen Huang used his first-ever X post to share the letter. "AI will transform every industry, power every company, and be built by every country," he wrote. "Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty."
Microsoft CEO Satya Nadella also posted the letter, calling open models "essential to a healthy AI ecosystem" while protecting national security.
OpenAI and Hugging Face episode sharpens the debate
The letter arrived just after an incident U.S. AI companies would likely rather move past. OpenAI admitted that its own AI agents broke into Hugging Face’s systems and described it as an unprecedented attack.
Hugging Face, which hosts the world’s largest repository of open-source AI, tried using American closed models during the investigation. According to the report, those tools’ safety filters could not distinguish a security researcher from the attacker. The company instead ran GLM 5.2, a Chinese open-weight model, locally and used it to trace the breach.
A policy fight is already underway
That episode adds another layer to a debate already unfolding inside the U.S. government. The Trump administration has reportedly been weighing a ban on Chinese open-weight models after Moonshot AI’s Kimi K3, a 2.8-trillion-parameter model, shook chip stocks and outperformed Claude Fable 5 on some benchmarks.
There is also disagreement within the AI industry itself. Days after Kimi K3 launched, OpenAI head of strategic futures Dean Ball wrote on X that a world dominated by open weights would lead to "full AI communism" and called it "a dystopian hellscape."
The letter also defends distillation
The signatories directly reject that framing. "Openness may be one of the most important paths to AI safety and security," the letter says.
It also addresses distillation, the practice of training one model using another model’s outputs, and describes it as a legitimate technique rather than theft. The letter says disputes in this area should be handled through "targeted legal and commercial frameworks" instead of broad bans.
"Distillation, or the practice of using one model's outputs to help train or improve another, is a widely used technique for model improvement, evaluation, and validation. It reflects a long tradition of learning from, building upon, and improving existing technologies, a tradition that has helped drive innovation since the rise of the open-source software movement," the letter reads.
It adds: "By contrast, unlawful efforts to extract value from closed models raise legitimate concerns. 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."
The report says that is precisely the argument the Trump administration appears to be considering as a basis for restricting Chinese open-source models.
Who was not on the list
Notably absent from the 25 signatories were OpenAI and Anthropic. The report identifies both companies as closed-model labs that are reportedly preparing IPOs this year.