“One is we iterate within Anthropic. We train the model, we're not happy with it, and we change the constitution. I think that's good to do.” — Dario Amodei
“The second level of loop is different companies having different constitutions. I think it's useful. Anthropic puts out a constitution, Gemini puts out a constitution, and other companies put out a constitution.” — Dario Amodei
Amodei discusses the challenges of governing powerful AI systems, including biosecurity risks, authoritarian misuse, and the need for new governance architectures that preserve freedom.