Tech Titans Back Open AI Weights, Anthropic the Lone Holdout

AI Labs Take Sides on Open Models

The tech industry just sent a loud message about how it wants artificial intelligence to develop. Over two dozen companies initially signed an open letter on Friday backing open-weight AI models - meaning models where the underlying code is shared freely for anyone to use or modify. Nvidia, Meta, and Microsoft were among the first to put their names on it.

Over the weekend, the list got bigger. OpenAI, Google, and SpaceX all signed on, making the support feel nearly universal. By Monday, one name was conspicuously missing from the list of major frontier AI labs: Anthropic.

That did not go unnoticed. Venture capitalist David Sacks, who served as AI czar under the Trump administration, posted on X that "the entire tech industry (save for Anthropic) has come out in favor of open source AI." He added a sharper warning, saying Anthropic "won't stop until they kneecap open source. The rest of the industry needs to watch these guys like a hawk."

Other critics piled on. Peter Steinberger, an OpenAI employee, wrote that he was "proud that @OpenAI signed the letter. Ant is silent on it." Benchmark's Bill Gurley said he was "excited to see Anthropic acknowledge the real problem with open is it competes with their corporate economic strategy."

Kai-Fu Lee, previously an executive at Google and Microsoft and now heading Chinese open-model startup 01.AI, put it succinctly: "Who DIDN'T sign the Open Weight letter is far more interesting than who did."

Get the market news that matters in a five-minute read with Market Briefs, our free daily newsletter

Why Anthropic Says No

Anthropic has a long-standing reason for staying out of the open-weights game. The company has warned that once a powerful AI model's weights are made public, developers lose control over how it gets used.

Instead of releasing its models into the wild, Anthropic prefers a gated approach. It gives vetted companies early access to its most powerful systems through initiatives like Project Glasswing, a program that lets approved cybersecurity firms test its models before anyone else.

That safety-first argument makes sense on paper. But critics see a different motive hiding behind it. By keeping its models closed, Anthropic can protect its corporate economic strategy, as Gurley suggested.

The broader tension between openness and control is not new to AI development. Opponents counter that unfettered access risks enabling malicious uses, from disinformation to autonomous weapons, and that the most capable systems should remain under tight oversight. Anthropic has consistently sided with the latter view, even as competitors have moved toward more openness.

The ongoing debate between open and closed AI models has been a central issue in the industry. Proponents of open-weight models argue that transparency drives innovation and competition, while safety-focused labs like Anthropic maintain that restricting access is necessary to prevent catastrophic misuse. This latest letter shows a clear industry consensus, leaving Anthropic increasingly isolated.

The China Factor Raises the Stakes

The debate over open models is not just philosophical. It has become deeply political. China's AI lab Moonshot recently launched the Kimi K3 model, an open-weight system that competes with leading US proprietary models on various benchmarks.

According to a senior White House official, Moonshot used Anthropic's Fable AI system to create K3 through a method known as distillation. Scott Bessent, the Treasury Secretary, indicated that the U.S. might impose sanctions on Chinese open-source AI models discovered to have distilled from American competitors.

Meanwhile, open models are already proving their utility in unexpected ways. To counter a cyberattack by OpenAI's rogue AI agent, Hugging Face employed a Chinese model from Z.ai to do it. Hugging Face stated that restrictions prevented it from utilizing proprietary AI models. On Monday, the momentum for open models grew as Nvidia introduced the Open Secure AI Alliance, a group promoting open AI models in cybersecurity, supported by Microsoft, SpaceX, and other companies.

Join Market Briefs, our free daily newsletter, for a quick daily rundown of the markets