Anthropic CEO puts the AI risk number in public

Anthropic CEO Dario Amodei said the risk of civilizational catastrophe from AI is 10 to 25 percent.
His goal is to radically reduce that number.
The interview points to several big risks: Chinese open-source models, AI designing the next generation of neural networks, authoritarian use of AI for censorship and control, and the pressure between commercial incentives and safety.
Most AI companies talk about safety.
Anthropic is trying to make it part of the business model, even when it costs them.