A coalition of AI and technology companies has urged US policymakers not to impose broad restrictions on open-weight models amid concerns about intellectual property theft and advanced Chinese AI systems. Signatories to the open letter include Hugging Face, Meta, Microsoft, Mistral, Nvidia, IBM, Palantir and Replit.
Open-weight models allow developers to download their underlying parameters and operate or customise them on independent infrastructure. The companies argue that this access supports competition, research, cybersecurity and lower-cost AI deployment.
Letter Distinguishes Distillation From Theft
The coalition warned policymakers not to treat established model-development methods as intellectual property misappropriation. It described distillation, where one model’s outputs help train or evaluate another system, as a widely used industry technique.
The letter acknowledged that companies may unlawfully extract value from closed models. However, it called for targeted legal and commercial action against proven misconduct rather than restrictions covering open models or distillation generally.
Replit Chief Executive Amjad Masad said banning Chinese open models could effectively restrict the wider open-model ecosystem. He noted that developers sometimes build new systems using several open models from different countries.
Cybersecurity Access Forms Part of Argument
The signatories also rejected claims that open weights are inherently unsafe because they can be used for cyberattacks. They argued that defenders need access to models with capabilities comparable to those available to attackers.
Hugging Face recently said commercial frontier models had blocked parts of its investigation into a cyberattack because their safety controls could not distinguish defensive work from malicious requests. The company instead used an open-weight model from Chinese developer Z.ai to analyse the incident.
The letter said wider access allows more researchers to discover vulnerabilities, test safeguards and develop defensive tools. It also called for investment in shared datasets, evaluation systems and computing access for researchers and start-ups.
Chinese Models Drive Policy Debate
The statement follows reports that the Trump administration has considered restrictions or sanctions involving advanced Chinese models. White House officials have accused Moonshot AI of using Anthropic’s Fable model to develop Kimi K3, although supporting evidence has not been publicly released.
OpenAI and Anthropic have previously warned that Chinese companies may be extracting capabilities from their proprietary models. Neither company signed the original letter, while Google DeepMind was also absent from the listed signatories.
The coalition asked policymakers to preserve competition by avoiding early restrictions that could push AI research and development outside the United States. It also encouraged wider access to computing infrastructure and common resources needed to train and evaluate models.
Featured image credits: Flickr
For more stories like it, click the +Follow button at the top of this page to follow us.