Jensen Huang breaks X silence to push open-weight AI models in Washington

Jensen Huang, the CEO of Nvidia, made his debut on X to endorse a high-profile industry letter championing open-weight AI models. The coalition backing the statement includes Microsoft, Meta, Hugging Face, and more than two dozen other organizations. Their central argument: open-weight models strengthen security, accelerate innovation cycles, and empower enterprises to retain sovereignty over their own AI systems rather than outsourcing entirely to cloud-hosted alternatives.

The move carries particular weight given Washington's current deliberations over potential curbs on certain Chinese AI systems. Despite the Trump administration initially framing open models as a strategic asset for U.S. competitiveness, regulatory scrutiny has intensified—signaling a more complex calculus inside the corridors of power.

A hybrid future, not an either-or#

Rather than viewing open and proprietary models as competitors, the letter reframes them as complementary tools suited to different jobs. Organizations increasingly download open-weight models, deploy them on private infrastructure, customize them for specific use cases, and keep sensitive data off third-party platforms. Nvidia itself has mapped this vision, where local and frontier models divide labor according to cost, speed, and control requirements.

The vendor lock-in concern underpins much of the argument. As enterprises make substantial AI investments, they want assurance against becoming captive to a single vendor or losing the capabilities they've developed. This preoccupation already shapes real business deals—Anaconda's recent purchase of Kilo, for instance, was explicitly framed around enterprises' reluctance to depend entirely on one AI provider.

Security through openness, not opacity#

Signatories reject the notion that closed systems are inherently safer. The letter states that proprietary models can be compromised, weaponized, or fail in ways that remain invisible to outside scrutiny. Concentrating advanced AI behind a handful of closed platforms magnifies systemic risk rather than reducing it. Palantir and Nvidia have already put this reasoning into practice, deploying open Nemotron models inside air-gapped government networks.

Distillation under the microscope#

Model distillation—the technique of transferring knowledge from a larger model into a smaller, more efficient one—has become a flashpoint in U.S.-China AI competition. The White House alleged that Moonshot AI distilled Anthropic's Fable model to create Kimi K3, a claim Moonshot denied. The letter contends that distillation should be treated as legitimate research rather than intellectual property theft, a position that has concrete implications: any new restrictions could reshape how enterprises build and deploy smaller models on-premises, edge devices, or cost-constrained infrastructure.