Nvidia, Microsoft urge US to avoid broad restrictions on open AI models

Twenty-four technology firms, led by Nvidia, Microsoft, and Meta, have asked Washington to keep its hands off open-weight AI models. The petition is a lobbying maneuver dressed as a policy debate, and the underlying motive is straightforward: the signatories sell the picks and shovels of the open ecosystem, and broad restrictions would shrink their addressable market.

The letter's timing is the tell. It landed the same week the White House raised concerns about Moonshot AI's Kimi K3 model and Treasury Secretary Scott Bessent floated sanctions against firms that improperly distill proprietary systems. The signatories are not arguing in the abstract. They are drawing a perimeter around a specific business model before regulators close it.

The coalition's framing is deliberately strategic. Open models, they contend, accelerate innovation, strengthen cybersecurity, and preserve American leadership. Each of those claims maps neatly onto a constituency that policymakers care about: startups want diffusion, defense contractors want security, and elected officials want geopolitical wins. The letter is engineered to give regulators a reason to say yes without forcing them to defend a position that looks soft on China.

The conspicuous absences matter more than the signatures. OpenAI and Anthropic, the two firms with the most to lose from open-weight competition, declined to sign. Google and xAI stayed out as well. The result is a coalition that looks broad but is actually narrow: it represents the hardware, cloud, and open-source segments of the AI stack, while the closed-model frontier labs sit on the sidelines. That alignment reveals where the real commercial pressure points are.

Nvidia's involvement is the most telling signal of all. Jensen Huang made his first-ever post on X to amplify the letter, a platform choice that underscores how seriously the chipmaker takes the regulatory threat. Nvidia's revenue depends on developers running models on its hardware, and open-weight systems are the fastest-growing category of that workload. Restrictions would not just slow competitors; they would constrain the demand curve for Nvidia's accelerators.

The labor-market implications are indirect but real. Open-weight models lower the capital barrier for startups building AI applications, which in turn expands hiring for engineers, researchers, and product teams outside the handful of frontier labs. If Washington tightens the rules, that hiring pipeline narrows, and the talent concentration around a few closed-model companies deepens. The signatories are not merely defending a technical preference; they are defending the labor market structure that their business models require.

The administration's next move will hinge on whether it treats distillation as an intellectual property issue or a national security one. The letter offers a convenient off-ramp: targeted legal remedies for bad actors, no sweeping restrictions on the technique itself. That framing lets policymakers appear tough on Chinese competitors while preserving the open ecosystem that American chipmakers and cloud providers depend on. It is the compromise the signatories are buying with this coordinated push.