It's Not Magic, a Sixth Street podcast A Conversation with Daniela Amodei, Co-Founder and President of Anthropic
18 snips
Feb 19, 2026 Daniela Amodei, co-founder and president of Anthropic and leader in AI safety and the Claude models. She discusses responsible scaling, rapid capability leaps in models, the role of regulation and human-in-the-loop safeguards, why generalists and specialists need each other, preserving culture during fast growth, and San Francisco’s role in tech and civic responsibility.
AI Snips
Chapters
Books
Transcript
Episode notes
Publish Data And Run Pilots To Drive Inclusive Adoption
- Publish transparent data and reports to surface adoption gaps and inform policy.
- Anthropic runs public reports and pilots (Gates Foundation, Clinton Health) to guide beneficial deployments into developing regions.
Training Ethics With A Written Constitution Improves Models
- Constitutional AI gives models a moral framework by training on documents like the UN Declaration of Human Rights and Apple's ToS.
- That approach improved both ethical behavior and overall model intelligence, per Daniela.
Invest In Modern Benchmarks To Measure Safety And Capability
- Build and adopt shared benchmarks and evaluations to judge safety and capability progress.
- Anthropic invests in internal and academic benchmark development because existing tests are outdated for modern LLMs.
