Short answers, pulled from the story.
Amanda Askell is a Scottish philosopher and AI researcher who has led the personality alignment team at Anthropic since 2021. Her work focuses on training the Claude model to exhibit positive character traits and developing Constitutional AI, a method for training AI systems using guiding principles and AI feedback. She appeared on the Time 100 AI list in 2024.
Amanda Askell left OpenAI after concluding that the company was not placing enough emphasis on AI safety. She had joined as a Research Scientist on the policy team in November 2018 and co-authored the GPT-3 paper before departing.
Constitutional AI, or CAI, is a method for training AI systems using a set of guiding principles rather than extensive human oversight. The model evaluates its own outputs against those principles and adjusts them accordingly, with AI feedback as the training signal. Askell has been a key contributor to CAI and is the primary author of the latest version of Claude's constitution, released in January 2026.
A 2023 study co-authored by Askell and Deep Ganguli found that the capacity for moral self-correction in large language models emerged at 22 billion parameters. Using three experimental benchmarks, the study showed that natural language instructions substantially reduced biased outputs in models of sufficient scale. Larger models could learn normative concepts like stereotyping and discrimination from training data without being given explicit definitions.
Amanda Askell studied philosophy and fine art at the University of Dundee. She received a BPhil in Philosophy from the University of Oxford and a PhD in Philosophy from New York University in 2018. Her doctoral thesis, titled Pareto Principles in Infinite Ethics, argued that ranking worlds containing infinitely many agents under certain plausible axioms generates contradictions for a wide range of ethical theories.
Claude's constitution provides the Claude model with guiding principles that allow it to evaluate and adjust its own responses, forming the basis of Anthropic's Constitutional AI approach. The latest version was released in January 2026. Amanda Askell is its primary author and is responsible for the majority of its text.