PeopleofSuperintelligence

Lab leader

Deep Ganguli

As of October 2026, Deep Ganguli is Director of Research at Stanford HAI. Their papers from 2021 onward examined alignment, predictability and harm reduction in large language models. In 2022 they were listed first on studies of predictability and of red teaming. In 2024 they jointly led Collective Constitutional AI, a study of aligning a language model with public input. A 2023 ICML workshop listed them as an Anthropic speaker.

Works
36

Publication numbers from OpenAlex, as of 15 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

  1. December 2021

    A laboratory for alignment

    'A General Language Assistant as a Laboratory for Alignment' is published in December 2021 with them among its authors, alongside Amanda Askell and Yuntao Bai.

  2. June 2022

    Predictability and surprise

    The 2022 ACM FAccT conference in Seoul presents 'Predictability and Surprise in Large Generative Models' in June 2022, with them listed first among its authors.

  3. August 2022

    Red teaming for harm reduction

    Published in August 2022, 'Red Teaming Language Models to Reduce Harms' lists them first among its authors.

  4. February 2023

    Moral self-correction in models

    'The Capacity for Moral Self-Correction in Large Language Models' is published in February 2023, with them listed first among its authors.

  5. 2023

    Talk on deployable generative AI

    At the ICML 2023 workshop 'Challenges in Deployable Generative AI', they give an in-person talk listed under Anthropic.

  6. June 2024

    Collective Constitutional AI

    'Collective Constitutional AI: Aligning a Language Model with Public Input' appears at the ACM FAccT conference in June 2024. They jointly lead the work with Saffron Huang, Liane Lovitt and Divya Siddarth.

  7. As of October 2026

    Director of Research at Stanford HAI

    They are HAI's Director of Research and will share highlights from its grant programs at a faculty town hall.

Sources and credits

Photo: We have no published photo of Deep Ganguli yet. Send us a free-licensed photo.

  1. Paper page - A General Language Assistant as a Laboratory for Alignment huggingface.co. Checked 15 October 2026.
  2. [2202.07785] Predictability and Surprise in Large Generative Models ar5iv.labs.arxiv.org. Checked 15 October 2026.
  3. Predictability and Surprise in Large Generative Models | Cool Papers - Immersive Paper Discovery papers.cool. Checked 15 October 2026.
  4. Paper page - Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned huggingface.co. Checked 15 October 2026.
  5. Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned | Cool Papers - Immersive Paper Discovery papers.cool. Checked 15 October 2026.
  6. Paper page - The Capacity for Moral Self-Correction in Large Language Models huggingface.co. Checked 15 October 2026.
  7. The Capacity for Moral Self-Correction in Large Language Models | Cool Papers - Immersive Paper Discovery papers.cool. Checked 15 October 2026.
  8. ICML Talk: Deep Ganguli, Anthropic icml.cc. Checked 15 October 2026.
  9. [2406.07814] Collective Constitutional AI: Aligning a Language Model with Public Input ar5iv.labs.arxiv.org. Checked 15 October 2026.
  10. Collective Constitutional AI \ Anthropic anthropic.com. Checked 15 October 2026.
  11. HAI Town Hall for Faculty | Stanford HAI hai.stanford.edu. Checked 15 October 2026.