Nvidia and Meta push back against U.S. restrictions on open AI models

The signatories, which include IBM, Palantir, Mistral, and the Linux Foundation, contend that open-weight systems are essential for developers to customize software and maintain control over internal data security. Rather than imposing blanket bans, the coalition advocates for targeted enforcement against intellectual property theft. Jensen Huang, CEO of Nvidia, emphasized the necessity of a dual-track ecosystem where both frontier closed models and open-source tools coexist to foster cybersecurity and cross-industry growth. Elon Musk has also publicly aligned with the push, signaling a rare consensus among industry leaders.

This debate intensifies as the Trump administration weighs potential sanctions against Chinese developers accused of unauthorized model distillation—using outputs from U.S. systems to train competing products. Treasury Secretary Scott Bessent has signaled a willingness to pursue bad actors while maintaining support for legitimate open-source development. However, the coalition warns that labeling standard technical practices like distillation as theft could inadvertently cripple the very research tools that drive American progress. With Chinese models such as Moonshot AI’s Kimi K3 climbing technical benchmarks, the industry remains wary that aggressive regulation may do more to hinder domestic developers than to curb foreign rivals.