Nvidia (NASDAQ: NVDA), Microsoft (NASDAQ: MSFT), Meta (NASDAQ: META) and 22 other organizations urged U.S. policymakers Friday to protect open-weight AI models from broad restrictions.
The coalition’s open letter, published July 24, says open-weight models are systems that anyone can download, look over, change and run on their own infrastructure. Signatories include Andreessen Horowitz, Hugging Face, IBM, The Linux Foundation, Mistral, Palantir, Perplexity and Y Combinator.
“Our AI leadership will be judged not by one frontier AI model, but by whether the United States builds a strong, open ecosystem that diffuses into every sector,” the coalition said.
The letter says open weights can let startups, established businesses, universities and public institutions use advanced AI without having to build models from scratch or pay frontier-model prices for each task. This allows organizations to reserve the largest models for the hardest problems and use cheaper, specialized systems for routine work.
The signatories also tied open weights to competition across model development, cloud infrastructure, chips and applications. Downloadable models can let customers retain control of their data and deploy systems across their preferred infrastructure, reducing dependence on a single vendor.
Security risks remain central
The coalition acknowledged that developers lose control of model weights after release, while modified versions may be difficult to trace or reverse. It argued that prohibition would still be the wrong response because cybersecurity defenders need systems comparable to those available to attackers.
“In fact, openness may be one of the most important paths to AI safety and security,” the letter said.
Closed models can also be breached or misused, according to the document, while failures may remain hidden from outside researchers. The signatories said open systems permit broader benchmarking, evaluation, red teaming and development of safeguards based on demonstrated harms.
The coalition called on Washington to expand compute access for researchers and startups, fund shared training resources, support domestic application development and avoid rules that could push AI work overseas. It also asked policymakers to distinguish common development methods from unlawful extraction of value from proprietary models.
Chinese models shape the policy debate
The letter comes as U.S. officials consider how to address American use of Chinese open-weight models. Nearly 200 Silicon Valley companies separately asked the Trump administration not to restrict that access, according to POLITICO. Startups told the publication that inexpensive Chinese models can substitute for costly usage credits from U.S. providers.
Administration officials have discussed measures involving Chinese AI laboratories and companies hosting their models, although concerns about suppressing U.S. innovation previously halted those efforts, Axios reported.
This week, Blockspace reported on The director of the White House Office of Science and Technology Policy Michael Kratsios posting on X his disapproval of Chinese firms “covert industrial distillation aimed at stealing proprietary U.S. technology.” The post followed news of Kimi K3’s distillation of new frontier models, such as Claude’s Fable 5.
The new letter focuses on distillation, which uses one model’s outputs to train or improve another. The coalition described it as a standard technique for improvement, evaluation and validation, while recognizing that unlawful extraction from closed systems raises legitimate concerns.
“Those concerns should be addressed through targeted legal and commercial frameworks rather than sweeping restrictions on techniques that play an important role in AI innovation,” the coalition said.