Open-letter coalition on open-weight AI tops 270 signatories

A web page presents the open letter "Open Weights and American AI Leadership", dated 24 July 2026, together with its list of signatories. The page states that more than 270 companies and organizations had signed as of 3 August 2026, among them Amazon, Google, Meta, Microsoft, NVIDIA, OpenAI, IBM, Mozilla and Hugging Face. The letter argues that American AI leadership depends on a strong, open ecosystem that spreads through every sector, rather than on a single frontier model.

The letter describes open weight models as ones that anyone may obtain, examine, alter and operate on their own systems. It says they widen access for startups, businesses, universities and public institutions, let organisations match a model to a task and cost, and spur competition among models, chips, cloud services and applications. It adds that they give customers more control over their data and help them avoid dependence on one provider.

The signatories acknowledge that open weights carry distinct risks, since released weights leave the developer's control and altered versions are hard to trace. They argue the answer is not prohibition: defenders facing attackers who use advanced AI need comparable models, and open models spread defensive capability and allow flaws to be found by many teams. The letter holds that openness could rank among the key routes to AI safety and security, and that depending only on closed models creates single points of failure.

For policymakers, the letter proposes wider compute access for startups and researchers, investment in shared datasets, tools and evaluation frameworks, and avoiding premature limits on open models. It also asks officials not to confuse distillation, a common technique for improving models, with misappropriation, and to address unlawful extraction from closed models through targeted legal and commercial tools.