Tech Giants Urge Washington to Keep Open-Weight AI Unrestricted

Open-weight AI models can be downloaded, inspected, modified and deployed on private infrastructure, lowering costs and reducing dependence on a single vendor. The policy debate has intensified as Washington weighs the security implications of increasingly capable Chinese models. Supporters argue that open systems broaden access, competition and cybersecurity research, while critics warn that released weights cannot easily be recalled or controlled, making the issue a test of how the US balances national security with technological leadership.

Nvidia, Meta, Microsoft, IBM, Palantir, Hugging Face and Andreessen Horowitz were among 25 organizations that published “Open Weights and American AI Leadership” on July 24, 2026. The letter urged US policymakers to avoid premature restrictions that could curb competition or push innovation overseas, and called for wider access to computing and shared research resources without specifying a funding amount. OpenAI and Anthropic did not sign, underscoring a widening policy split between open-model backers and leading closed-model developers.