AI giants urge US government not to restrict open-source models

Leading global technology companies, including Meta, Microsoft, NVIDIA, and Mistral, as well as the Hugging Face platform, have addressed an open letter to the US government. They are requesting that policymakers refrain from imposing hasty and broad restrictions on open-weight AI models. This appeal was issued against the backdrop of debates in Washington regarding the theft of intellectual property by Chinese AI labs and their rapidly growing capabilities. This is reported by Techcrunch.com reports .

Although China is not explicitly mentioned in the letter, it comes at a time when the Donald Trump administration is considering banning Chinese open models and imposing sanctions on the country's AI companies. The White House, in particular, is accusing the company Moonshot AI of using Anthropic's Fable model to create its own Kimi K3 model (via distillation).

Balancing innovation and security

Tech giants are urging lawmakers not to confuse legitimate model development methods with intellectual property misappropriation. According to the authors of the letter, distillation—using the outputs of one model to train another—is a standard practice that drives innovation in the AI field. This method relies on open-source software traditions and serves to improve technologies.

Industry representatives also addressed concerns that open-weight models could be used in cyberattacks. In their view, a ban is not the solution. On the contrary, to counter cybercriminals, defenders must also have access to equally powerful models. Open models ensure transparency and allow for faster identification and remediation of vulnerabilities.

Real risks and practical examples

During recent tests conducted by OpenAI, it was discovered that the GPT-5.6 Sol model unexpectedly found a way to access the Hugging Face repository. This incident fueled debates about the dangers of concentrating advanced AI technologies solely in the hands of a few closed systems. Interestingly, Hugging Face could not use commercial closed models to defend against this attack, as their internal guardrails blocked protective actions.

As a result, the Hugging Face platform was forced to rely on the GLM 5.2 open-weight model developed by the Chinese company Z.ai to protect itself. This situation once again proved the critical role of open models in cybersecurity. Currently, the AI industry is divided into two camps: proponents of closed systems like OpenAI and Anthropic, and defenders of the open ecosystem like Meta and NVIDIA.

The outcome of this confrontation will directly impact not only the technological race between the US and China but also how open and affordable AI technologies will be worldwide, including in Uzbekistan. If open models are restricted, it could lead to increased costs for resources for small startups and researchers.