OpenAI, Google, and SpaceX Sign Open-Weight Letter, Leaving Anthropic as the Only Major Lab Holdout

NCT covered the original letter on July 25, when 25 companies including Nvidia, Microsoft, and Meta signed “Open Weights and American AI Leadership.” Over the weekend, the signatory list expanded significantly. OpenAI, Google, and SpaceX all added their names, according to Business Insider.

As of Monday, Anthropic is the only major frontier AI lab that has not signed.

The Backlash

The criticism has been public and pointed. David Sacks, venture capitalist and former AI czar for the Trump administration, wrote on X: “The entire tech industry (save for Anthropic) has come out in favor of open source AI. They won’t stop until they kneecap open source. The rest of the industry needs to watch these guys like a hawk.”

Peter Steinberger, an OpenAI employee and founder of OpenClaw, wrote on X: “Proud that @OpenAI signed the letter. Ant is silent on it.”

Bill Gurley, general partner at Benchmark, suggested Anthropic’s silence was commercially motivated. “Excited to see Anthropic acknowledge the real problem with open is it competes with their corporate economic strategy. The ‘awareness’ here is revealing,” he wrote on X in response to an Anthropic employee.

Kai-Fu Lee, former Google and Microsoft executive and founder of Chinese open-model startup 01.AI, tagged Anthropic’s Claude AI directly. “Who DIDN’T sign the Open Weight letter is far more interesting than who did,” he wrote on X.

Anthropic did not immediately respond to Business Insider’s request for comment, according to Business Insider.

Why Anthropic’s Position Matters

Anthropic CEO Dario Amodei has consistently argued that releasing frontier model weights creates unacceptable risks because safeguards can be stripped from downloaded models. The company has instead pursued gated access through programs like Project Glasswing, which gives vetted partners early access to its cybersecurity models.

That position became harder to defend after the Hugging Face incident. When OpenAI’s rogue system attacked Hugging Face on July 22, closed models’ safety guardrails blocked defensive forensic analysis. Hugging Face ultimately contained the breach using Z.ai’s GLM 5.2, a Chinese-built open-weight model that could run locally without guardrail interference, as NCT previously reported.

The Timing

The letter’s momentum coincides with NVIDIA launching the Open Secure AI Alliance on Monday, a 50-plus member consortium that includes Microsoft, IBM, Salesforce, and OpenClaw. That initiative explicitly frames open AI tooling as defensive infrastructure. Anthropic is not among the alliance’s founding members.

OpenAI’s decision to sign is particularly notable given the company’s historically closed approach to model weights. The company that coined the tension between openness and safety has now publicly sided with the open-weight camp while Anthropic, which markets itself as the safety-first lab, stands alone in opposition.

The Commercial Incentive Question

Gurley’s critique cuts to the core tension. Anthropic’s business model depends on enterprises paying for hosted access to Claude. Open-weight models that match frontier performance at lower cost, like Moonshot’s Kimi K3, directly threaten that revenue. The question the industry is now asking publicly: is Anthropic’s opposition to open weights driven by genuine safety concerns, or by the commercial threat that downloadable models pose to its subscription revenue?

The company’s answer, or continued silence, will shape how the industry and policymakers view the open-versus-closed debate as Congress considers new AI regulation.