Leading AI companies including Hugging Face, Meta, Microsoft, Mistral and Nvidia have urged policymakers to avoid broad, premature limits on open-weight AI models. In an open letter, the group argues that innovation should not be slowed by sweeping rules while governments assess how to respond to concerns around model development and intellectual property.
The appeal arrives as debate intensifies in the United States over how to handle rapidly advancing AI systems from China. Rather than targeting open-weight models as a category, the signatories say regulators should focus on specific misuse cases and use targeted legal and commercial tools where needed.
The letter also defends distillation, a common technique in which one model helps train or improve another. According to the companies, this method has long supported progress in AI and should remain part of the ecosystem that drives research, testing and validation.
Another key point is cybersecurity. The letter says open models can strengthen defense by giving researchers and security teams access to comparable capabilities for detecting, simulating and responding to emerging threats. That openness, it argues, can improve transparency and help vulnerabilities surface faster across many teams.
The discussion reflects a broader split in the AI sector. Some closed-model developers favor tighter controls as open systems become more capable and widely available, while infrastructure providers and open-model advocates see broader access as a path to faster experimentation, more competition and wider adoption.
As AI policy evolves, the balance between openness, safety and innovation could shape the next phase of global technology development.
