Nvidia and Microsoft lead open weight AI plea to Washington

This letter is the clearest articulation yet of an “open weight bloc” inside the AI industry, and it lands at a moment when Washington is openly weighing bans on Chinese open models. Nvidia, Microsoft and Meta are not acting as altruists here, but they are spelling out a strategic view of the ecosystem: that US leadership depends less on one unbeatable closed API and more on a thick layer of downloadable models that developers, enterprises and allies can run anywhere. In that frame, trying to shut down open weights looks like a way to kneecap your own industrial base while doing little to slow adversaries.

For the race to AGI, the stakes are obvious. If regulators squeeze open weights too hard, experimentation and red teaming will migrate to jurisdictions that are more permissive, and a lot of the frontier talent that today sits on Llama and Kimi style stacks will simply go elsewhere. If, on the other hand, the US codifies a path where frontier closed models and powerful open weights coexist, then much of the world’s AI infrastructure, from chips to tooling, will still be tied to US aligned platforms.

What makes this different from past open source advocacy is who is talking. When Nvidia, Microsoft, Palantir, Mistral, Hugging Face and a16z all sign the same document, they are telling policymakers that open weights are not a hobbyist side show but a pillar of their own business plans.