AIBRIEF
Nvidia, Meta, Microsoft, IBM, Palantir, and 20 others asked Washington not to restrict downloadable AI model code, four days after a reported push to ban Chinese models.
Twenty-five companies, including Nvidia, and the Linux Foundation, published a three-page "Open Weights and American AI Leadership" letter on July 24 asking Washington not to restrict "open-weight" models: downloadable model code anyone can run and fine-tune, in contrast to a closed service like ChatGPT. OpenAI, Anthropic, and Google did not sign.
The letter landed four days after the Trump administration was reported to be reviving a push to ban Chinese models such as DeepSeek. The document itself names no country or lab. It landed while Washington is weighing its next move.
The asks: expanded compute access for startups and researchers, public investment in shared datasets and evaluation frameworks, and a stance that distillation (training one model's outputs on another) should not be treated as misappropriation of closed models. All five model developers on the list (Meta, Mistral, Black Forest Labs, Arcee AI, Reflection) already publish their weights; three signatories (a16z, Y Combinator, Emergence Capital) are venture funds that back open-weight startups.
Nvidia CEO Jensen Huang posted the letter, writing in his first X post: "The world needs both frontier closed models and frontier open models." Unite.ai and Decrypt read the move as a defensive plea against broad restrictions; ZeroHedge called it a "commoditization cartel" going to Washington.
What remains unknown: whether the absence of the three closed-API frontier labs is read as consensus, or as the fight the letter is trying to preempt.