25 Tech Companies Urge U.S. Policymakers Not to Restrict Open-Weight AI Models

Twenty-five technology companies have joined forces to urge U.S. policymakers not to restrict open-weight AI models.

The coalition, which includes Microsoft, NVIDIA, Meta, IBM, Palantir, Hugging Face, Dell, ServiceNow, and Mozilla, released the letter as Chinese AI developers continue to narrow the gap with leading U.S. systems.

In the letter titled Open Weights and American AI Leadership, the coalition argues that open-weight AI should follow the same path as open-source software. Signatories say broader access to AI models would strengthen U.S. innovation while giving organizations more flexibility in how they develop and deploy AI.

The letter argues that organizations should have greater control over the AI systems they use. It also makes the case that transparency and proprietary development can coexist, while urging policymakers to address misuse through targeted enforcement instead of broad restrictions.

OpenAI and Anthropic were not among the signatories. Their absence was notable given their influence over the direction of frontier AI development and policy.

Why It Matters: Organizations are beginning to decide which AI platforms they’ll build around for years to come. If policymakers place new restrictions on open-weight models, enterprises could have fewer options for running AI on their own infrastructure, customizing models with proprietary data, or choosing alternatives to proprietary AI services.

• Open-Source Software Sets the Example: The coalition argues that AI is at a moment similar to the early days of open-source software.Technologies like Linux became the foundation for industries ranging from cloud computing to scientific research because developers could study the code and adapt it to new uses. Signatories believe open-weight AI can create the same kind of foundation by making advanced models available to more organizations.

• Access Opens the Door to More Innovation: Organizations wouldn’t have to train models from scratch or pay frontier-model prices for every workload. Instead, they could choose the model that best fits the task while reserving frontier-scale systems for the most demanding work. The letter argues this makes AI more economically sustainable as adoption spreads across industries and encourages competition throughout the AI market, from model developers to cloud providers and software vendors.

• Organizations Keep More of What They Build: Businesses should be able to decide where AI models run and keep sensitive data on infrastructure they control, while adapting models with their own proprietary knowledge. Over time, organizations retain the value created from those investments instead of becoming dependent on a single provider.

• Security Doesn’t Depend on Closed Models: Open-weight models introduce real risks because they can be modified after release, a point the coalition openly acknowledges. The letter argues that transparency also makes it easier for researchers and security teams to identify vulnerabilities and improve defenses over time. It warns that relying on only a small number of proprietary AI systems carries security risks of its own.

• Policy Should Encourage Growth While Addressing Abuse: According to the coalition, policymakers should expand access to AI compute along with shared datasets, development tools, and evaluation frameworks that support innovation. The letter also distinguishes legitimate model distillation from intellectual property theft, arguing that unlawful behavior should be addressed through targeted legal and commercial action instead of broad restrictions on open-weight AI.

Go Deeper -> Nvidia, Microsoft, Meta warn against ‘premature restrictions’ of open-weight models – CNBC