Nvidia, Microsoft, Meta and Palantir joined more than 20 technology companies and organizations in urging US policymakers not to impose broad restrictions on open weight artificial intelligence models.

The coalition published an open letter Friday arguing that models whose weights can be downloaded, inspected, modified and operated on private infrastructure are essential to US technological leadership and wider AI adoption. Other signatories included IBM, Andreessen Horowitz, Hugging Face, Mistral, Mozilla, Perplexity and Y Combinator.

OpenAI, Anthropic and Google, three of the leading developers of closed frontier models, did not sign the letter.

The appeal follows increased scrutiny of Chinese AI developers after Beijing based Moonshot AI released Kimi K3. The model’s capabilities and lower operating costs intensified debate in Washington over whether Chinese open models should face restrictions because of national security and intellectual property concerns.

Michael Kratsios, director of the White House Office of Science and Technology Policy, said the US had information indicating that Moonshot used distillation to develop Kimi K3 from Anthropic’s Fable model, according to the Financial Times. Distillation involves using the output of a stronger model to train or improve another model.

Treasury Secretary Scott Bessent said the Trump administration supports open source AI but warned that sanctions and Entity List designations could be considered when Chinese companies conduct covert industrial distillation that crosses into intellectual property theft.

The companies behind Friday’s letter argued that policymakers should distinguish ordinary distillation from unlawful efforts to extract value from closed models. They called for suspected misconduct to be addressed through targeted legal and commercial measures rather than broad restrictions on open models or commonly used training methods.

The letter said open weight models allow businesses, universities and public institutions to use advanced AI without developing models from scratch or paying frontier model prices for every task. It also argued that open models increase competition and reduce dependence on a small group of providers.

The signatories acknowledged that open models carry risks because developers lose control over the weights after release. However, they argued that greater access allows researchers to examine model behavior, identify vulnerabilities and develop safeguards.

Nvidia Chief Executive Jensen Huang separately defended Chinese open models this week, saying US companies should be allowed to use them. Huang argued that cheaper and more accessible models would increase AI adoption and ultimately drive greater demand for chips, data centers and computing infrastructure.

The debate has created a widening divide between companies that benefit from open model availability and frontier developers whose business models depend on maintaining control over their most advanced systems.