Anthropic Publishes Position on Open-Weights Models: No Ban, But Calls for Distillation Crackdown

What happened

Anthropic published "Our position on open-weights models" on anthropic.com/news on July 27, 2026 — a direct response to criticism of Anthropic's absence from the open-weight AI coalition led by Nvidia, Meta, Microsoft, and others.

Context and impact

Anthropic is the only major AI lab that did not sign the open-weight AI letter. CEO Dario Amodei explained why — he is not opposed to open-weight models per se, but identifies specific security risks. This statement is key for understanding Anthropic's position in the current geopolitical debate.

Details

• Anthropic has never advocated for banning open-weights models
• Non-dangerous open-weight models are a "public good" per Amodei
• Main concerns: authoritarian governments developing more powerful AI than the US; misuse for cyberattacks and bioweapons
• Proposed solutions: chip export restrictions to China + crackdown on state-backed industrial-scale distillation
• Mandatory safety testing for all sufficiently capable models, open and closed
• Key distinction: protecting chips from authoritarian access is more effective than banning models