AI Policy

25 AI Firms Ask Washington to Back Open Models

Twenty-five US technology firms signed a letter asking Washington to avoid premature restrictions on open-weight AI. Here is what it means for owners in Europe who want off vendor lock-in.

AI PolicyBy Servola Tech Desk2026-07-243 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

25 AI Firms Ask Washington to Back Open Models

What the 25 companies actually signed

The clearest signal was not a paragraph in the letter but who wrote it. Jensen Huang, Nvidia's chief executive, joined X for the first time to back the document, titled 'Open Weights and American AI Leadership' and hosted on Microsoft's corporate-responsibility site. Alongside Nvidia sat Microsoft, Meta, Andreessen Horowitz, IBM, Dell, Palantir, Mistral, Hugging Face and Y Combinator, 25 signatories in all. Two names were absent, and they are the two that sell only closed frontier models: OpenAI and Anthropic.

The asks are concrete. The group wants Congress to expand compute access for startups and universities, fund shared training datasets and evaluation frameworks, and hold off on what it calls 'premature restrictions' on open models. The timing is not an accident. On July 16 the Beijing lab Moonshot released Kimi K3, an open-weight model strong enough to reset what a free download can do, and Washington is now arguing over whether to curb open models, especially Chinese ones.

Why an owner in Frankfurt reads this differently

In the United States the fight is about leadership and speed. In Europe it is already about paperwork. From August 2 the AI Act's obligations for general-purpose models apply, and a model that crosses the systemic-risk compute threshold has to document its training data, run evaluations, and report serious incidents, open weights included. A US letter that says 'do not restrict' and an EU rulebook that says 'disclose and test' are not the same policy, and any owner deploying an open model on both sides of the Atlantic now lives under both.

The practical read is not political. An open-weight model is the one procurement lever that survives a vendor's price rise or a change of terms: you can run it in a Frankfurt or Dublin data centre, keep your data off a US cloud, and fix your cost per million tokens instead of renting it monthly. That is the same sovereignty argument European buyers have been making for two years, now being made in Washington by the firms that sell the chips.

The commercial bet under the letter

Follow the incentives and the letter reads as positioning, not principle. Nvidia, Microsoft, Dell and the cloud names make more money the more models exist to run, because every open model someone fine-tunes is more hardware rented and more inference sold. A world of two closed giants sells fewer chips than a world of a thousand open ones. That is why the chip and cloud layer signed and the two closed-model labs did not.

The owner's takeaway sits underneath the lobbying either way. Open weights lower your switching cost no matter who wins in Washington or Brussels. The question worth asking your vendor this quarter is not whether open models are safe in the abstract, but whether the model you depend on could be self-hosted if the contract turned against you. If the answer is no, the letter is describing a freedom you do not yet have.

Frequently asked questions

What is an open-weight AI model?

It is a model whose trained parameters are published for download, so anyone can run it on their own hardware and fine-tune it, rather than only calling it through a provider's API. Kimi K3, Meta's Llama line and Mistral's models are open-weight; OpenAI's GPT and Anthropic's Claude are not.

Does the letter change any law?

No. It is a lobbying document asking US lawmakers to expand compute access and avoid new restrictions on open models. It has no force in Europe, where the EU AI Act's general-purpose model rules begin applying on August 2, 2026 regardless of what Washington decides.

Why did OpenAI and Anthropic not sign?

Both sell closed frontier models accessed through paid APIs, so a thriving open-weight ecosystem competes with their core product. The signatories are mostly firms that profit from more models existing: chipmakers, cloud providers and open-model labs.

The letter will be argued over for months. The leverage it points at, a capable model you can hold, run, and leave with, is available to owners now, and worth building into the next contract you sign.

Open weightsAI ActNvidiaMetaMicrosoftUS policyDigital sovereigntyKimi K3

13,524 Answers Show Where AI Goes Quiet

The Oversight Board tested 10 models from 6 providers across 10 jurisdictions. Refusal rates were 34% in restrictive countries versus 14% in permissive ones. The method is now a procurement instrument.

5 min read

Digital Sovereignty

Europe's Sovereign AI Now Runs on American Money

On 21 July 2026 Microsoft committed billions to expand Mistral's AI in Europe, on Nvidia chips, with an air-gapped deployment mode. Here is why sovereignty now means where your data sits, not who owns the stack.

4 min read

Law & Regulation

The EA Buyout Now Hinges on Brussels

The largest take-private in history has cleared EA shareholders and US antitrust. The last real gate is the EU - and not classic merger review, but the Foreign Subsidies Regulation, facing its first blockbuster test.

3 min read

← All articles