As the United States considers a response to Chinese AI, the industry is urging not to impose broad restrictions on open weight

Several AI companies, including Hugging Face, Meta, Microsoft, Mistral, and Nvidia, have signed an open letter urging policymakers not to impose broad “premature restrictions” on open-weight AI models. The letter comes as Washington debates how the United States should respond to allegations that Chinese AI labs are stealing intellectual property from their American counterparts and growing their capabilities.

The letter did not mention China at all, but comes on the heels of reports that the Trump administration is considering banning Chinese open-weight models, and possibly issuing sanctions against AI companies in the country. The White House even accused Moonshot AI of distilling Anthropic’s Fable model to train the recently released Kimi K3 model, which is impressive by all accounts.

The letter appears intended to discourage blanket bans on Chinese models, as well as ensure that the administration’s response to alleged Chinese distillation does not spill over into broader restrictions on open-weight AI or popular technologies like distillation:

Policy makers should be careful not to confuse legitimate model development techniques with embezzlement. Distillation, or the practice of using the output of one model to help train or improve another model, is a widely used technique for model improvement, evaluation, and validation. It reflects a long tradition of learning from, building on, and improving existing technologies, a tradition that has helped drive innovation since the advent of the open source software movement.

By contrast, illegal efforts to extract value from closed models raise legitimate concerns. These concerns should be addressed through targeted legal and business frameworks rather than blanket restrictions on technologies that play an important role in AI innovation.

Or as Replit CEO Amjad Massad (who also signed the letter) told TechCrunch: “I think banning Chinese open forms is as good as banning open forms in general.” He noted that Thinking Machines Lab’s new open model, Inkling, was trained with the help of Moonshot’s Kimi 2.5. “It’s an ecosystem, a precedent [a ban would] The group is bad.”

The letter also rejects industry arguments that open-weight models are inherently dangerous because they expand access to powerful models, which can be used for cyberattacks or other nefarious activity, without any oversight.

“The correct response to this risk is not to ban open weights. In a world where cybersecurity attackers use advanced artificial intelligence, defenders need access to models with similar capabilities so they can detect, simulate, and respond to emerging threats,” the letter said. “Open models expand defensive capability, increase transparency, and allow vulnerabilities to be discovered and remedied across many teams.”

Last week, OpenAI revealed that while testing GPT-5.6 Sol and another unnamed model, a system exploited a vulnerability in its testing environment to access a Hugging Face repository containing a solution to the coding standard. One could argue that the model’s target was not malicious, and that he was actually cheating on the test to get higher scores. But the incident has sparked debate about the dangers of concentrating advanced AI technology behind a handful of closed providers.

Hugging Face said it was unable to defend itself against the attack using commercial border AI models because its own guardrails hampered its efforts. The closed AI models I used were not able to differentiate between being asked to create an exploit for an attacker and being asked by a defender to discover it. Instead, the company was forced to focus on using Chinese AI company Z.ai’s GLM 5.2, a powerful open-weight model, to defend itself against the attack.

The letter highlights the division in the AI ​​industry. Companies such as OpenAI and Anthropic have urged the administration to respond to alleged intellectual property theft by Chinese AI companies as open-weight models rapidly grow in capabilities. The outcome could have major implications for their business models, which are threatened by the proliferation of cheap, highly capable and easily accessible AI models.

These companies, along with closed-source AI developers like Google DeepMind and SpaceX, are conspicuous by their absence at the bottom of this letter.

Those who signed the letter have a clear economic interest in seeing open AI models flourish. Companies like Nvidia, Microsoft Azure, and other infrastructure providers have a vested interest in pushing toward commodity models: if models are interchangeable, people will buy more GPUs, rent more cloud capacity, and build more applications.

The letter encourages policymakers to expand access to computing for startups and researchers; Investing in shared training assets such as datasets, tools and assessment frameworks; And “[keep] Multi-boundary by avoiding premature restrictions on open models that stifle competition or push innovation outward.

This article has been updated with commentary from Amjad Massad, CEO of Replit.