Dario and Daniela Amodei on Claude's values, Pentagon red lines and why they left OpenAI

Anthropic co-founders Dario AmodeiDealroom has a profile for this one. Try Dealroom → and Daniela Amodei sat down with Walter Isaacson for The Builders (published 27 October 2026, interview taped 6 July 2026).

The GPT-2 moment Dario called GPT-2 "the first real test of the scaling laws hypothesis" and said: "I personally was super freaked out by it... you just scale this thing up another factor of a million and it'll be better than humans at everything." He said the same dismissal now greets Anthropic's plan to develop drugs for neglected rare diseases.

Training Claude on principles, not rules Dario said early rule lists produced a "spiky shape" of around 170 arbitrary rules, and that teaching Claude basic principles "allows Claude to generalize to new situations". Daniela argued sycophancy is harder to stamp out than dishonesty or harm, and said that because Anthropic has "committed to not put ads in the models" it has no incentive to keep users talking longer. Dario described stress tests he compared to CIA vetting, and mechanistic interpretability as "a bit like giving the model an X-ray or a brain scan".

OpenAI, trust and the Pentagon On leaving OpenAI, Dario said "honesty is one of our values" and that without good-faith disagreement "how can you possibly hope to resolve them? That's why we left." He noted all seven co-founders are still at the company. On autonomous weapons, he cited reliability and accountability concerns, warning of "one person having their finger on a button that controls a million robots or a million drones", and said efforts in Congress now aim to enact Anthropic's red lines into law.

Source The Builders with Walter Isaacson: Dario & Daniela Amodei (YouTube)