Nvidia, Meta, and Microsoft Tell Washington: Don’t Kill Open-Source AI

Twenty-five companies just told Washington to leave open source AI alone. Nvidia, Microsoft, Meta, Andreessen Horowitz, Hugging Face, IBM, Dell, and 18 others signed a letter published on Friday called. “Open Weights and America’s AI Leadership.” Their argument: banning open source models will not protect the US, it will only give the AI ​​market to a few closed labs.

Open weights are AI systems whose “weights” – the billions of numbers that decide how the model will respond – are freely published. Anyone can download them, run them on their own servers, and modify them. This is in contrast to closed models such as OpenAI’s GPT or Anthropic’s Claude, which only bypass the company’s paid software or API.

The letter draws a direct line to the open programming movement of the 1980s, arguing that AI has reached a similar fork in the road. “Our AI leadership will not be judged by one type of AI, but by whether the United States builds a robust, open ecosystem that spans every sector,” the letter reads.

Nvidia CEO Jensen Huang shared the letter in his first letter at X. “AI will change every industry, power every industry, and be built by every country,” he wrote. “Open models strengthen security and cybersecurity, accelerate innovation and expansion, and enable autonomy.” Microsoft’s Satya Nadella echoed it, calling open models “essential to a healthy AI ecosystem” and defending national security.

There comes a time when America’s AI industry will be forgotten. OpenAI admitted that their AI assistants were involved Hugging Face systemsthey are calling it an unprecedented attack. When Hugging Face, which hosts the world’s largest open source AI platform, tested closed American models for analysis, the device’s security filters would not tell the security researcher from the attacker. So the company ran GLM 5.2, China’s open source version, instead — and used it to track down the breach.

This news is disrupting the fight that is already going on in the state. Trump administration says it will try to ban Chinese light models after Moonshot AI’s Km K32.8-trillion-parameter model, chip stocks and beat Claude Fable 5 on other indicators. Not everyone in the AI ​​industry wants to be open to success, either. A few days after Kimi K3’s launch, OpenAI’s head of strategic futures, Dean Ball, wrote on X that an open world with burdens leads to “total AI communism” and called it a “dystopian hellscape.”

The letter pushes back on the design directly: “Openness can be one of the most important steps towards the safety and security of AI.” It also draws a line around distillation, the practice of training one model using the output of another, calling it a legitimate method rather than theft that deserves “legal and commercial methods” rather than restrictions.

“Distillation, or the practice of using the output of one model to help train or refine another, is a widely used technique in model control, evaluation, and validation.” It reflects a long tradition of learning, building, and improving existing technologies, a tradition that has helped drive innovation since the rise of the open source software movement,” the letter says.

“Contrary to this, illegal efforts to extract value from closed models raise legitimate concerns. Those concerns should be addressed through revised regulations and businesses rather than restricting processes that play a vital role in AI innovation.”

This is the conflict that the Trump administration seems to be in measure to justify the ban on China’s open brands.

None of the 25 signatories: OpenAI and Anthropic, two closed labs both said to be planning IPOs this year.