Industry letter argues for open weights in American AI policy
A July 2026 industry letter hosted by NVIDIA argues that open-weight models support access, competition, user control, and security research while carrying distinct risks.
The letter makes a policy case for open-weight AI, including broader access, competition, customer control, and the ability of researchers and defenders to inspect and strengthen models.
It also acknowledges that released weights can be difficult to trace or reverse and calls for targeted policy rather than broad restrictions. The signatory list includes NVIDIA and many AI, cloud, software, and infrastructure organizations.