The artificial intelligence landscape is heating up again. Last Friday, Jensen Huang, CEO of Nvidia, published his first post on X to share a letter titled Open Weights and American AI Leadership, a three-page document co-signed by 25 companies. Signatories include giants like Nvidia, Microsoft, Meta, IBM, Dell Technologies, Palantir, and Hugging Face. The letter asks the U.S. administration to avoid what it calls "premature restrictions on downloadable AI models", at a time when the Trump administration is considering a ban on Chinese AI models, such as Moonshot AI and DeepSeek. Notably missing from the list are OpenAI, Anthropic, and Google, an absence that has not gone unnoticed.
The 25 signatories represent the entire AI ecosystem
The list of signatories spans chipmakers, server vendors, cloud operators, enterprise software firms, security companies, and venture funds. Names include Dell, Microsoft, IBM, Box, ServiceNow, CrowdStrike, Telnyx, Replit, Perplexity, and funds Andreessen Horowitz, Y Combinator, and Emergence Capital. Model developers that already publish open weights, such as Meta, Mistral, and Black Forest Labs, are naturally present. The Linux Foundation also joined, which stewards the OpenMDW-1.1 license used by Nvidia to release Nemotron 3 Ultra in June, a 550-billion-parameter model. The letter comes in response to events just days earlier: Treasury Secretary Scott Bessent stated on Fox Business that the administration would examine Chinese open-source models for potential intellectual property theft, threatening sanctions. Huang replied to Axios that American firms should be allowed to use Chinese models, calling accusations of backdoors a "misconception".
Sponsored Protocol
Meanwhile, the chip sector is experiencing tensions: Qualcomm announced a 10% price increase for Snapdragon chips starting September 2026, a sign that technological competition is intensifying on the hardware front as well.
Sponsored Protocol
Distillation and compute access: the letter's hot points
The letter addresses two critical issues. First is distillation, the practice of training one model on another's outputs. The signatories ask policymakers not to treat distillation as misappropriation, arguing that unlawful extraction from closed models should be handled through targeted legal frameworks rather than broad limits on the technique. The second point concerns compute access: the letter calls for expanded compute access for startups and researchers, alongside public investment in shared datasets and evaluation frameworks. Huang wrote in his X post: "The world needs both frontier closed models and frontier open models." He also revealed that one in every four tokens generated today comes from an open model. Downloadable weights are served from enterprise clusters, regional clouds, and on-premises racks, not just a handful of hyperscaler API endpoints. This creates an alternative market that does not depend on internal TPU or Trainium programs.
Sponsored Protocol
The letter has drawn mixed reactions: on one hand, open weights advocates see a step forward for transparency and digital sovereignty; on the other, absent companies like OpenAI and Anthropic may fear that open models could be misused. It remains to be seen whether Washington will heed the appeal or proceed with a ban on Chinese models, as anticipated by the Trump team. The game of AI regulation has only just begun.
