Microsoft Is One of 25 Tech Giants to Back Open-Weight AI, but Anthropic Is Not

In Brief

Twenty-five American tech companies, including Microsoft, Meta, Nvidia, IBM, and Palantir, published a joint letter urging policymakers not to restrict open-weight AI models. 

Nearly every major AI leader lined up behind it. Anthropic is the lone frontier lab standing against it.

What Happened

Dozens of major American tech companies signed a joint letter titled Open Weights and American AI Leadership, urging policymakers not to restrict open-weight AI models, which are models anyone can download, inspect, modify, and run on their own infrastructure rather than access only through a provider. The letter came just as Washington debates banning Chinese AI models.

The letter argues America's AI leadership "will be judged not by one frontier AI model, but by whether the United States builds a strong open ecosystem that diffuses into every sector." (A frontier model is one of the handful of most powerful AI systems in the world.) Open weights, it says, expand access to the AI economy, strengthen competition, give customers control over their data and models, and even improve safety, because "relying solely on closed models is not inherently safe" and concentrating advanced AI in a few frontier labs creates single points of failure. It also defends distillation, the common practice of training a new model on a bigger model's outputs, warning policymakers "not to conflate legitimate model development techniques with misappropriation."

Then the endorsements poured in. Nvidia CEO Jensen Huang used his first-ever post on X to share the letter, writing that "the world needs both frontier closed models and frontier open models." Microsoft CEO Satya Nadella called open-weight models essential to a healthy AI ecosystem, OpenAI CEO Sam Altman said he wants the US to win in AI in both open source and proprietary models, those accessible only through the company that makes them, and Google CEO Sundar Pichai was "very happy to support this on behalf of Google."

Mark Zuckerberg called open source "a positive important force for both empowering people and preventing centralization." Elon Musk voiced support as well. Noticeably absent among the endorsers was Anthropic. Anthropic researcher Julian Schrittwieser mocked Microsoft's newfound openness, joking he could not wait for the open sourcing of Windows and Microsoft Office.

On Episode 226 of The Artificial Intelligence Show, SmarterX founder and CEO Paul Roetzer, who ranked the letter as possibly the most important AI development of the year, explained why it exists and why Anthropic is not endorsing it. 

The Key Numbers

25 - American tech companies and organizations behind the letter

1st  - Post ever on X from Nvidia CEO Jensen Huang, which he used to share the letter

300+ - Millions of downloads of Google's open-weight Gemma models

12 to 18  - Months Google typically lags behind before open-weighting a former frontier model

3  - Years since Dario Amodei's Senate testimony warning about open models

Why Everyone but Anthropic Signed

Every signature has a business case behind it. Roetzer's first question was why Microsoft, one of OpenAI's biggest backers, would champion open models at all. "You always have to step back and say, why would a different organization support this? What is their stake in this?" he says. For Microsoft, the answer is Azure, its cloud business. The more AI gets used anywhere in the economy, open or closed, the more companies spend on Microsoft's cloud. So their business does not depend on selling proprietary models, Roetzer explains.

The timing is deliberate. Roetzer sees three forces behind the letter arriving now:

1. Chinese labs, including the one behind Kimi K3, keep launching near-frontier models with far fewer resources

2. The distillation fight needed a public defense, and the industry wants to shape regulation before Washington acts.

3. Google's support fits its playbook: for roughly three years it has released open-weight versions of its former frontier models 12 to 18 months after they debut, keeping only its most powerful model closed. Google DeepMind CEO Demis Hassabis added that "a strong and secure open system is important for the world to benefit from AI," pointing to 300 million+ downloads of its open Gemma models.

Anthropic is trading popularity for its safety argument. "Anthropic very clearly is standing on their own island right now," says Roetzer, and it is "losing a lot of what my kids would call aura points right now in the AI industry." The stance traces back to CEO Dario Amodei's July 2023 Senate testimony. Amodei told senators that "in most scientific fields, open source is a good thing" and that there is room for models on the smaller and medium side. His concern was the frontier: "When you control a model and you're deploying it, you have the ability to moderate usage," Amodei testified. "When a model is released in an uncontrolled manner, there's no ability to do that. It's entirely out of our hands." He warned frontier models would become seriously risky within two to three years. That was three years ago.

Roetzer himself is not fully in either camp.

"I'm still conflicted personally on this. I get the benefits of open weights. I truly do. And I don't dispute those at all.

But I feel like some of the accelerationists just throw aside the possibility that maybe it just doesn't go right. Maybe we can't keep up with these agents as they go out into the wild if you give open weights or open source to the most powerful frontier models."

— Paul Roetzer, founder and CEO of SmarterX, Episode 226 of The Artificial Intelligence Show

SmarterX Take

For company leaders, this fight has significant implications for their businesses. Whether powerful open-weight models stay available in the U.S. shapes what companies can run on their own infrastructure, what AI costs, and how much control they keep over their data. The letter pitches open weights as the way every organization matches the right model to the right job at the right cost, and that flexibility is exactly what enterprise AI strategies are starting to depend on.

Microsoft's support is principled and also good for Azure. Anthropic's dissent looks self-serving to critics and also matches three years of public safety warnings. Both can be true at once.

What to Watch

Washington's response will set the boundaries of the open AI economy. The letter exists to influence regulation, and the debate over banning Chinese models is the near-term test. How policymakers treat open-weight releases and distillation in the coming months will show whether the plea worked.

A single high-profile disaster could flip the politics overnight. If a powerful model is misused in a way that causes serious damage, the government could feel forced to act, and act forcefully. That risk is part of why the industry is making its case now, before it's forced into it.

Are Leaders Confident Evaluating AI?

Only 49% of professionals rate their confidence in evaluating AI-powered technology as high or very high, according to the 2026 State of AI for Business Report. The largest group, 42%, sits at medium confidence. The open versus closed fight will decide which models companies can choose from, but most of the people making those choices still feel only moderately equipped to evaluate them.

The full report, based on 2,100+ responses from professionals across roles, functions, and industries, maps where organizations actually stand on adoption, governance, training, and tooling. Read the full report →