Jensen Huang’s Open Weights Letter Doubles to 50 Signatories in One Day

On July 24, 2026, Nvidia CEO Jensen Huang published his first-ever post on X, sharing a letter called "Open Weights and American AI Leadership" that urges Washington not to restrict downloadable AI models. The letter launched with 25 company signatures. Within a single day, that number doubled to 50, with OpenAI, Google, AMD, Cisco, Cloudflare, GitHub, Block, and Ollama among the additions. Two notable names are still absent: Amazon and Anthropic. The speed of coalition-building and the identity of the holdouts both carry real implications for businesses buying AI services today.

What happened

Detail	Fact
Date of Huang’s X post	July 24, 2026
Original signatories	25 companies
Signatories after one day	50 companies
Huang’s post views	11 million
Notable new signatories	OpenAI, Google, AMD, Cisco, Cloudflare, GitHub, Block, Ollama
Notable absences	Amazon, Anthropic

Jensen Huang used his first post on X to share “Open Weights and American AI Leadership,” a letter asking the US government not to place restrictions on AI models that can be downloaded and run locally (known as open weights models). At launch, OpenAI had not signed. That detail was widely reported. By the following day, OpenAI was among roughly two dozen companies that had added their names after the letter was already public.

The letter now represents chipmakers, cloud providers, security vendors, application companies, and the venture firms that back them. Two hosted copies reportedly showed different signature counts on the same afternoon, which gives you a sense of how fast the roster was moving.

Why the two absences matter more than the 50 signatures

Amazon is Anthropic’s largest investor. Anthropic trains and runs its models on Amazon’s Trainium hardware, among other infrastructure. Google is also an Anthropic backer and still signed the letter. That makes Amazon’s decision to stay out the more deliberate-looking of the two.

Neither company has offered a public explanation. The possible readings range from straightforward commercial interest (Anthropic’s business is built on selling closed frontier model access) to a genuine safety position that open weights, once released, cannot be recalled. Anthropic has argued the second point publicly for years. Both motivations can coexist.

If you currently buy AI services from Amazon or Anthropic, their absence is a reasonable thing to raise at your next vendor review. Not because they are necessarily wrong, but because their position on open weights policy will shape what you are actually allowed to run on their platforms in the future.

What does “open weights” actually mean for your business?

The letter’s central concept is sovereignty: the ability to keep operating if a vendor changes its terms. Open weights address one half of that question and leave the other half open.

• Model inspection and retraining: You can download the weights, examine how the model works, and fine-tune it on your own data. That is the concrete benefit.
• Chip access: Weights without compute are useless. Access to the GPUs and accelerators needed to run large models is governed by export policy and allocation queues, not by whether the model weights are public. Nvidia, which organized the letter, operates at this layer.
• Data residency: Running a model on your own infrastructure turns data location from a contract clause into an architectural reality. That matters for regulated industries.
• Agent auditability: When an AI agent approves a payment, accesses a customer record, or takes an action on your behalf, can you prove it was authorized? Can you reconstruct what version of the model was running, and whether it was the one your security team approved? Open weights leave this question entirely untouched.

The letter itself acknowledges the limitation. It concedes that once weights are public, the original developer loses control, and altered copies are difficult to trace. The proposed safeguard is that outside researchers will catch problems over time. That may be true at an industry level. It is not an answer a CIO can give a regulator who wants to know who checked a specific transaction.

What does the speed of this coalition tell us?

The letter did not build quietly over months before going public. It launched, Huang’s post reached 11 million views, and signatures arrived while coverage was still being written. Two dozen companies joined after publication.

For any business weighing whether to take a public position on an emerging policy question, that is the new timeline. The gap between a position becoming visible and becoming crowded is now measured in hours, not weeks. Companies that wait for consensus to form before deciding where they stand are already late.

This dynamic is worth watching in the context of broader US government AI policy debates, where proposed legislation is moving faster than most corporate legal teams anticipated.

Our take

The open weights letter is largely a lobbying document, and the speed of its growth tells you more about coalition mechanics than about the underlying technical question. Fifty signatures in a day is impressive optics. It does not resolve the genuine tension between model openness and accountability.

The Anthropic and Amazon absences deserve more attention than they have received. Anthropic has been the most consistent major lab voice arguing that open release of frontier models carries irreversible risk. That is a serious position backed by a body of published safety research, not just a business preference. It may be both, but it should be engaged on the merits rather than dismissed because the other 50 companies disagree.

For clients thinking about their AI infrastructure choices, the honest framing is this: open weights give you more control over the model layer, and that matters. But the harder governance problems, specifically who authorized what action, which model version ran in which environment, and how you prove it, sit at the application and agent layer. Open weights do not solve those. If you are building AI into workflows, this is exactly where a proper AI integration approach earns its cost, because model access and operational accountability are two different problems.

What to do about it

1. Check whether your current AI vendors (especially Amazon Web Services and Anthropic) have published a position on open weights policy. If not, ask your account team directly.
2. Map which of your AI-driven workflows involve agent actions (payments, data access, communications) and document the authorization chain for each.
3. Review your data residency requirements. If regulations require data to stay in a specific jurisdiction, open weights models you host yourself may reduce your compliance exposure compared to API-based services.
4. Watch whether Amazon or Anthropic issue a public statement. Their explanation, if it comes, will clarify whether this is a commercial or a safety position, and that distinction matters for how you evaluate their roadmap.

The letter is a starting point for a policy conversation, not a solution to the governance questions your AI systems already face today.

Source: Bing News · OpenAI