OpenAI quietly signs letter from Nvidia, Microsoft, and Meta warning about dangers of ‘premature restrictions’ on open-weight AI models — as the White House accuses China of stealing from Anthropic

• July 25, 2026

• Several AI companies in the US have published an open letter
• The letter is campaigning for unrestricted open-weight AI
• Meanwhile the US government accuses Moonshot AI of IP theft

Any hopes AI tech reporters may have had about easing quietly into the weekend have been dashed by some of the biggest names in the business sharing an open letter in support of open-weight AI models — while the White House has accused China of ‘large scale’ theft from US companies in building its open-weight AI.

Nvidia CEO Jensen Huang got the ball rolling with his first-ever tweet, linking to a statement championing open-weight AI models, those models that can be downloaded, viewed, and edited by anyone. Models like this are essential for making AI “more accessible, adaptable, and widely available” the open letter says.

It’s co-signed by a slew of well-known tech and AI companies, including Meta, Microsoft, Perplexity, and IBM. While ChatGPT developer OpenAI — which has had more of a closed-model focus, to date — didn’t appear on the original post from Huang, it has now also backed the sentiments.

Essentially, these companies are campaigning for openness in AI to drive innovation and access, while also arguing against ‘premature restrictions’ by governments. It’s no coincidence that the US government led by President Trump has recently tried to take a more prominent role in evaluating AI models and deciding when they should be released.

It’s also notable that Anthropic and Google are missing from the list of backers to this open letter. We don’t have any official comment yet from these companies, so it’s hard to judge the reasoning — Google regularly releases open AI models, after all — but it could be something to do with recent accusations aimed at Moonshot AI in China.

Open and closed

Part of the background to this open letter is that Chinese company Moonshot AI stands accused of distilling Claude’s (closed) Fable 5 model in order to build its own (open) Kimi K3. In simple terms, distilling works by asking millions of questions of an AI model, and using the responses as training data.

It’s something Moonshot AI has denied, while hyping up the potential of the open-weighted, freely downloadable Kimi K3 AI model, which is being released on Monday — another factor in this open letter coming out now. The US government is calling it IP theft, and has issued some heavy-handed responses.

“We support open-source AI and the innovation it unlocks,” US Treasury Secretary Scott Bessent posted on X. “But open source is not open season on American IP. When PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table.”

What Nvidia, Meta, Microsoft, and the other open letter co-signers seem to be doing is trying to get ahead of any potential clampdown by the US government on open-weighted models. While everyone agrees that these AI models are beneficial, it’s not clear what the next regulatory steps might be.

However, the open letter is vague and general enough that you can read between the lines in a variety of different ways. It may be intended to allude to the Moonshot AI case, or it may not. What’s clear is that we’re coming to a crunch point for AI regulation and innovation across the globe, and everyone has a perspective.

Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.

Several AI companies in the US have published an open letter The letter is campaigning for unrestricted open-weight AI Meanwhile the US government accuses Moonshot AI of IP theft Any hopes AI tech reporters may have had about easing quietly into the weekend have been dashed by some of the…