Something unusual happened in the AI industry this week. Fifty-eight technology companies, including fierce competitors like NVIDIA, Microsoft, Meta, Google, and OpenAI, put their names on the same piece of paper.
The document is called “Open Weights and American AI Leadership.” It was published on July 24, 2026, and it is essentially a message to Washington: do not restrict access to open-weight AI models.
NVIDIA CEO Jensen Huang used his first-ever post on X to amplify the letter, writing that open models “strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.” The list of co-signatories reads like a who’s-who of the AI industry: AMD, Andreessen Horowitz, Cohere, Hugging Face, IBM, Mistral, Y Combinator, Palantir, Replit, and Perplexity, among many others.
Why This Is Happening Now
The letter did not emerge in a vacuum. In the weeks before its publication, the Trump administration was reported to be considering a ban on Chinese AI models, particularly after the US government accused Moonshot AI of using Anthropic’s Fable model to build Kimi K3. Treasury Secretary Scott Bessent had signalled that sanctions were possible.
The coalition’s letter does not mention China, Moonshot, or DeepSeek directly. But the timing is not coincidental. The signatories are drawing a clear line: before Washington moves to restrict which AI models American businesses can use, it needs to think carefully about the unintended consequences for the broader AI ecosystem.
What the Letter Actually Says
The coalition’s core argument is built around four claims.
First, open-weight models allow organisations to build AI applications without training a model from scratch. This lowers the barrier to entry considerably, especially for smaller businesses and regulated industries that cannot afford the compute costs of frontier model training.
Second, openness drives competition. When many organisations can build, adapt, and deploy advanced models, costs fall and benefits spread more widely than they would in a closed ecosystem controlled by a handful of providers.
Third, user control matters. Businesses using open-weight models can keep their data on their own infrastructure, adapt the model to their specific needs, and deploy it wherever their operations require. That is a meaningful advantage for industries with strict data sovereignty requirements.
Fourth, and perhaps most surprising to some observers, the letter argues that openness may actually be one of the most important paths to AI safety. When model weights are publicly available, a broad community of researchers can examine how the models behave, find vulnerabilities, and push for improvements. Closed systems make that kind of community scrutiny impossible.
The letter also draws a direct parallel to the open-source software movement of the 1980s. The open-source community’s work now underpins most of the internet and is embedded in systems used by the US military and federal agencies. The signatories argue the same dynamic applies to AI model weights today.
The Holdouts
Two companies notably absent from the list are Amazon and Anthropic. Amazon has its own set of interests as a cloud provider and Anthropic investor. Anthropic, whose Fable model sits at the centre of the Moonshot controversy, may have felt it could not publicly advocate for open models at this particular moment.
The fact that OpenAI signed is arguably the bigger story. OpenAI has historically been cautious about open-weight releases, and its inclusion signals either a genuine shift in position or a calculated political move ahead of what looks like an increasingly active regulatory environment.
What This Means for Business
If you are running a business that uses AI tools, the outcome of this policy debate will affect your options in a real and practical way.
A government decision to restrict access to open-weight models would likely push more businesses toward a smaller number of closed API providers. That means higher costs, less flexibility, and greater dependency on the pricing and policy decisions of a handful of large companies.
The coalition’s position is that a more open ecosystem keeps costs competitive and puts more control in the hands of the businesses using AI, not just the companies selling it.
For businesses currently evaluating AI infrastructure decisions, this debate is worth following closely. The practical implications of what Washington decides here will be felt in licensing agreements, model availability, and infrastructure costs for years.
If your organisation is still figuring out what role AI should play across your operations, the strategic question is not just which model to use today. It is how to build an AI capability that does not leave you exposed when the rules change.
Enterprise DNA’s Omni services are built to help businesses deploy AI in a way that gives you real operational leverage, not just access to a chatbot. If that conversation is worth having, start here.