AI People Wiki

Builders & open source

Amanda Askell

Leads Anthropic's Character work (finetuning and AI alignment), Anthropic

  • Leads Anthropic's Character work (finetuning and AI alignment), Anthropic12

Amanda Askell is a philosopher working on finetuning and AI alignment at Anthropic, where Askell's team trains models to be more honest and to have good character traits. Anthropic credits Askell as the lead of its Character work and the primary author of Claude's constitution, released in January 2026 under a CC0 1.0 deed. Before Anthropic, Askell was a research scientist on OpenAI's policy team, working on AI safety via debate and human baselines for AI performance, and is a co-author of the GPT-3 and CLIP papers. Askell is also first author of Anthropic's 2021 paper "A General Language Assistant as a Laboratory for Alignment".123456

At a glance

Current role
Leads Anthropic's Character work (finetuning and AI alignment)
Company
Anthropic
Sector
Company
Education
NYU, Oxford
Work listed
3
Sources cited
9
Profile reviewed

Career

Amanda Askell's CV lists doctoral study in philosophy at New York University from 2011 to 2018, with a thesis on Pareto principles in infinite ethics advised by Cian Dorr, and an earlier BPhil at the University of Oxford with a thesis on objective epistemic consequentialism. Askell's philosophical work has centered on ethics, decision theory, and formal epistemology.71

From 2018 to 2021 Askell was a research scientist on the policy team at OpenAI, working on AI safety via debate and on human baselines for AI performance. Askell joined Anthropic in 2021 as a research scientist in alignment finetuning, and Anthropic now credits Askell with leading its Character work.712

Research

From 2019 to 2021, Askell's publications included work on AI safety and social science, release strategies for language models, and cooperation in responsible AI development, along with co-authorship of the GPT-3 paper in 2020 and the CLIP paper in 2021.8459

At Anthropic, Askell was the first author of "A General Language Assistant as a Laboratory for Alignment" in December 2021 and co-authored later papers including Constitutional AI (2022), work on moral self-correction in language models (2023), and studies of sycophancy and of specific versus general principles for constitutional AI (2023). Askell's team trains models to be more honest and to have good character traits, and develops finetuning techniques meant to scale to more capable models.681

Claude's constitution

Claude's constitution, published by Anthropic in January 2026, names Askell as its primary author. Its acknowledgements say Askell wrote most of the text and led its development through several rounds of revision, with significant contributions from Joe Carlsmith and others, and the document is released under a CC0 1.0 deed.23

Work and contributions

  • Claude's constitution Published research

    Primary author who wrote the majority of the text and led its development through multiple rounds of revision, according to the document's preface and acknowledgements; Anthropic released it under a CC0 1.0 deed.23

  • A General Language Assistant as a Laboratory for Alignment Published research

    First-listed author, marked as a core research contributor, of the 2021 Anthropic paper studying baseline techniques and evaluations for a helpful, honest and harmless text-based assistant.6

  • CLIP Open weights

    Co-author, with an OpenAI affiliation, of "Learning Transferable Visual Models From Natural Language Supervision" (2021), which releases code and pre-trained weights under the MIT license.59

Timeline

  1. Completed a PhD in philosophy at New York University and joined OpenAI's policy team.7
  2. Published "AI Safety Needs Social Scientists."8
  3. Co-authored "Language Models are Few-Shot Learners," the GPT-3 paper.4
  4. Joined Anthropic as a research scientist in alignment finetuning.7
  5. Co-authored the CLIP paper, Learning Transferable Visual Models From Natural Language Supervision.5
  6. First author of "A General Language Assistant as a Laboratory for Alignment."6
  7. Co-authored "Constitutional AI, Harmlessness from AI Feedback."8
  8. Anthropic published Claude's constitution, with Askell as its primary author.32

Other documented roles

Roles the sources document that are not confirmed as current.

  • Research scientist on the policy team, OpenAI15

Where they work

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

9 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. Amanda Askell — About MeAmanda Askell · Official page · accessed Oct 21, 2026
  2. Claude's Constitution (preface and acknowledgements)Anthropic · Official page · accessed Oct 21, 2026
  3. Claude's new constitutionAnthropic · Announcement · published Jan 22, 2026 · accessed Oct 21, 2026
  4. Language Models are Few-Shot Learners (arXiv 2005.14165)arXiv · Paper · published May 28, 2020 · accessed Oct 21, 2026
  5. Learning Transferable Visual Models From Natural Language Supervision (arXiv 2103.00020)arXiv · Paper · published Feb 26, 2021 · accessed Oct 21, 2026
  6. A General Language Assistant as a Laboratory for Alignment (arXiv 2112.00861)arXiv · Paper · published Dec 1, 2021 · accessed Oct 21, 2026
  7. Curriculum Vitae — Amanda AskellAmanda Askell · Official page · accessed Oct 21, 2026
  8. Publications and Preprints — Amanda AskellAmanda Askell · Official page · accessed Oct 21, 2026
  9. openai/CLIP repository (MIT License)OpenAI (GitHub repository) · Repository · accessed Oct 21, 2026