Anthropic CEO Dario Amodei published a public response Monday to claims that the company wants U.S. policymakers to ban open-weight AI models.

Amodei wrote that Anthropic has "never advocated for a ban on open-weights models" and said models without dangerous capabilities can be a public good for businesses, developers, and researchers. The post followed an industry letter defending open-weight AI and a broader debate over whether U.S. companies should face restrictions on Chinese open-weight systems.

The distinction in Amodei's post is between commercial use of existing open-weight models and the national security risks he says come from more powerful systems. He argued that bans on use by U.S. businesses would not address his main concerns, because the most serious risks involve authoritarian governments building frontier models, cyberattacks, biological misuse, and industrial-scale distillation.

Amodei said open-weight models can carry higher risk than closed systems once capabilities become dangerous, because released weights are difficult to monitor, guardrail, or withdraw. But he framed that as a reason for targeted measures, including tighter controls on advanced chips, chipmaking equipment, smuggling, and large-scale distillation.

The post gives Anthropic a clearer position in a fast-moving policy fight: it is not calling for a broad open-weights ban, but it is still pushing for limits around compute access and the most capable AI systems.