Nvidia and 24 Companies Defend Open-Weight AI in Washington

Jensen Huang Opens His X Account to Defend Open-Weight AI on 24 July 2026

Nvidia chief executive Jensen Huang made his first-ever post on X on 24 July 2026 — not to announce a new GPU, but to publish a policy letter co-signed by 24 other major technology companies. The letter, titled Open Weights and American AI Leadership, urges the United States government to avoid broad restrictions on open-weight AI models. Huang wrote that AI would transform every industry, power every company, and be built by every country, and that open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The post arrived as Washington was actively debating its response to evidence that Chinese AI laboratories had extracted capabilities from US frontier models through large-scale distillation, and the letter is a pre-emptive argument against legislation that would restrict open-weight model publication or use.

What the Letter Says and Who Signed It

The letter's central argument is that restricting open-weight AI models would concentrate capability inside a small number of closed-model vendors, reduce independent third-party safety research, and prevent governments, universities, and smaller businesses from building and controlling their own AI systems. The 25 co-signatories include Nvidia, Microsoft, Meta, Google, AMD, Hugging Face, IBM, Dell, Mozilla, Mistral, Cohere, Cloudflare, GitHub, CrowdStrike, Replit, Perplexity, and Y Combinator. The breadth of the list — spanning model providers, cloud infrastructure companies, developer platforms, and security vendors — is deliberate. It is intended to show that the open-weight ecosystem is not a minority position but the operational foundation of a large portion of the US AI industry, and that any restriction targeting Chinese open-weight models would have significant collateral effect on US-based development.

The Distillation Allegation That Triggered the Policy Debate

The letter responds directly to allegations that Chinese AI laboratories used distillation to extract capabilities from US frontier models. The White House stated that Moonshot AI used distillation on Anthropic's Claude Fable model to develop its Kimi K3 system. Distillation involves querying a target model at high volume to generate training data for a student model, effectively transferring learned behaviour without accessing the underlying weights or training data. Treasury Secretary Scott Bessent said in mid-July that the administration was examining leading Chinese open-weight models for evidence of IP theft and could impose sanctions if wrongdoing was established. The policy options under discussion include a ban on Chinese open-weight models in US federal procurement and potential export-control-style restrictions on sharing US open-weight model weights with Chinese entities.

Why OpenAI and Anthropic Did Not Sign

Two of the most prominent names in US AI are absent from the signatory list. OpenAI and Anthropic, whose frontier models were specifically named in the distillation allegations, did not sign the letter. Both companies have consistently argued that frontier-capability models carry sufficient dual-use risk to warrant access controls that open-weight release does not provide. Anthropic's safety frameworks are built on the principle that the most capable models should not be freely downloadable without training-time alignment applied by the developer. OpenAI has taken a similar position in its published safety commitments. Their absence signals that the open-weight debate within the US AI industry is unresolved, and that the legislative outcome will depend on how Washington balances the competitiveness argument from the coalition against the safety case advanced by the companies whose models were allegedly distilled.

What the Open-Weight Policy Debate Means for Indian Software Teams

For Indian software companies and AI product teams, the Washington open-weight debate has two practical dimensions. The first is access continuity. If US policy moves toward restricting open-weight model publication or use by entities in non-allied countries, the availability of models like Meta's Llama family, Mistral's models, and future US open-weight releases could become a policy variable rather than a stable technical constant. Indian teams that have built production infrastructure around self-hosted US open-weight models should monitor legislative progress closely, particularly as the debate moves from executive branch discussion to potential Congressional action. The second dimension is the Chinese model question. Federal procurement bans on Chinese AI models are unlikely to affect Indian commercial use directly, but the precedent of treating AI model weights as national security assets — in both Washington and Beijing — is a structural shift in how the international AI supply chain is governed. Indian teams building for US government-adjacent enterprise clients should evaluate whether their AI tooling carries exposure to models that could face downstream access restrictions in regulated procurement contexts.

The Bottom Line

On 24 July 2026, Nvidia CEO Jensen Huang used his first X post to co-publish Open Weights and American AI Leadership, a letter signed by 25 companies including Microsoft, Meta, Google, AMD, Hugging Face, IBM, and Mistral, urging Washington to avoid broad restrictions on open-weight AI. The letter responds to White House allegations that Chinese AI laboratories, including Moonshot AI, used distillation attacks on US frontier models — specifically Anthropic's Claude Fable — to build competing systems. Treasury Secretary Bessent has stated that sanctions are possible if IP theft is confirmed. OpenAI and Anthropic, whose models are at the centre of the distillation allegations, did not sign the letter. For Indian AI product and software teams, the debate introduces a near-term risk to open-weight model access continuity and establishes a precedent for treating AI model weights as national security assets on both sides of the technology competition.