Nvidia and Palantir ask US not to restrict open AI models

In short: US tech companies and investors are pushing back on ideas to restrict “open” AI models, after new fears about Chinese AI progress and alleged IP theft.

What's going on

A group including Nvidia, Palantir, Microsoft, Meta, IBM, venture firm Andreessen Horowitz, and others signed an open letter asking US policymakers not to place early limits on “open-weight” AI models. These are AI models that people can download, change, and run on their own computers or servers. Think of it like software you can install at home, instead of a service you must use through a company’s website.

The letter comes after louder calls to restrict Chinese AI models, especially following the release of Kimi K3 from Beijing-based Moonshot AI. US officials have raised concerns that Chinese labs could be using US technology in improper ways. Michael Kratsios, who leads the White House Office of Science and Technology Policy, said the US has information suggesting Moonshot used “distillation” (training a smaller model by copying the answers of a stronger model, like a student learning by watching an expert’s solutions) based on Anthropic’s latest model.

US Treasury secretary Scott Bessent said sanctions could be considered if Chinese firms cross the line into intellectual property theft. He also said the Trump administration supports open-weight AI.

What to watch

The US could still move toward restrictions on Chinese AI models or on certain training methods. At the same time, companies backing open models argue that keeping them available helps researchers find problems and add safeguards, since more people can inspect how they behave.

Source: Financial Times