A coalition of technology companies, investors and developer organizations is urging Washington to preserve open-weight artificial intelligence, arguing that U.S. leadership depends on more than a handful of closed frontier models.

The July 24 letter, titled “Open Weights and American AI Leadership,” defines open-weight models as systems whose parameters can be downloaded, inspected, modified and run on an organization’s own infrastructure. Signatories include Nvidia, Meta, Microsoft, Google, OpenAI, Amazon, Hugging Face and a long list of software and security companies.

The group argues that open weights lower cost, reduce dependence on a single provider and give researchers and defenders more ways to examine model behavior. It acknowledges that a released model is difficult to recall or control, but says broad restrictions would concentrate risk in a small number of opaque systems and push innovation elsewhere.

Openness is becoming a policy fault line

The letter arrives as U.S. officials weigh the security implications of powerful downloadable models and the role of Chinese developers in the global ecosystem. Closed-model companies have also warned that distillation can be used to reproduce capabilities from systems they spent heavily to train.

The signatories ask policymakers not to confuse legitimate distillation and model development with unlawful extraction. That is a commercially important distinction, but it will be difficult to draw in practice. The same technical method can support research, competition or misappropriation depending on the access, intent and contractual terms.

For enterprise buyers, the open-versus-closed debate is not ideological. Open weights can improve portability, customization and data control, but they transfer more responsibility to the operator. A company that runs a model itself must secure the environment, evaluate modifications, monitor use and maintain safeguards that a service provider might otherwise manage.

Closed systems offer another trade: less operational control in exchange for a vendor-managed stack and clearer accountability for some failures. Recent containment incidents have shown that closed does not automatically mean safe, just as downloadable does not automatically mean transparent to the average user.

Washington’s challenge is to regulate demonstrated risks without freezing the market around today’s largest firms. A policy that is too broad could remove a source of competition and defensive research. A policy that ignores capability could distribute systems whose misuse cannot be reversed.

The letter makes the industry’s preferred answer clear: build evaluation tools, compute access and targeted legal rules around an open ecosystem. The next question is whether government can make that ecosystem accountable without closing it.

Sources for editorial review

Drafting note: This draft was prepared with AI assistance from the linked source material and requires author review, independent fact-checking and final editorial approval before publication.