Our position on open-weights models

In a post addressing recent discussions, Anthropic CEO Dario Amodei clarifies that the company has never advocated for a ban on open-weights models, viewing safe open-weights models as a public good. Amodei outlines his primary national security concerns: authoritarian regimes developing superior AI for military dominance or repression, and the potential misuse of powerful AI for cyber or biological attacks. To address these threats, he advocates for restricting chip sales to China, cracking down on industrial-scale model distillation, and mandating rigorous pre-release safety testing for all sufficiently capable models, regardless of whether they are open or closed.

(Source：Anthropic)