Microsoft, Nvidia and Meta-backed letter urges U.S. support for open-weight AI

A coalition of major technology companies and AI ecosystem groups has published a policy letter urging U.S. officials to preserve support for open-weight artificial intelligence models, arguing that downloadable models are becoming a strategic part of the country's AI infrastructure.

The letter, titled Open Weights and American AI Leadership and dated July 24, says U.S. leadership should not be measured only by which company controls the most capable frontier model. Instead, the signatories argue that the country needs a broad ecosystem where startups, universities, public institutions and established businesses can inspect, adapt and run AI models on their own infrastructure.

Why the letter matters

Open-weight models expose trained model weights for download and local deployment. That makes them different from closed models delivered only through a hosted API. The letter says this openness can lower costs, reduce lock-in and give organizations more control over their data and deployment choices. It also frames open weights as a competition issue, saying that wider access can create rivalry across model developers, cloud providers, chips, applications and services.

The policy argument lands as Washington debates how to handle powerful downloadable models, including models developed outside the United States. The letter does not ask policymakers to ignore misuse. It instead calls for targeted legal responses to unlawful extraction or abuse, while warning that broad restrictions could push innovation overseas or concentrate too much power in a small number of closed systems.

Security is central to the pitch

The signatories make a security case as well as an economic one. They argue that defenders need access to capable models so they can test systems, simulate attacks and build safeguards, and that closed systems can also be breached, misused or fail in ways outsiders cannot evaluate. In that framing, openness is not presented as risk-free; it is presented as a way to widen scrutiny and resilience.

Reporting from Tom's Hardware said the group included Nvidia, Microsoft, Meta, IBM, Dell Technologies, Palantir, Hugging Face, Mozilla, the Linux Foundation and several other companies and investors. The outlet also noted that OpenAI, Anthropic and Google were not listed among the signatories. That split makes the letter notable: it puts chipmakers, enterprise software firms, open-model developers and infrastructure players on record together at a moment when AI policy is increasingly tied to national competitiveness, cybersecurity and control over computing platforms.

For businesses deciding whether to build on open or closed AI systems, the letter is not a product launch. It is a coordinated policy signal that open-weight models are likely to remain a major battleground in U.S. AI regulation, procurement and security planning.

Sources

• www.microsoft.com
• www.tomshardware.com

Cover image: North Charleston, source, licensed under BY-SA.