PeopleofSuperintelligence

Researcher

Paul Röttger

Also known as Paul Rottger

Paul Röttger is a researcher who studies the safety and societal impacts of large language models, and his datasets and test suites check how AI systems handle hate speech, values and human feedback. He completed a DPhil at the Oxford Internet Institute in 2023 under Helen Margetts and Janet Pierrehumbert, was first author of the HateCheck test suite in 2021, and co-founded the content moderation startup Rewire, which ActiveFence acquired in March 2023. After joining Dirk Hovy's MilaNLP Lab at Bocconi University as a postdoctoral researcher in 2023, he won an Outstanding Paper Award at ACL 2024 and coauthored The PRISM Alignment Dataset, which won Best Paper in the Datasets and Benchmarks track at NeurIPS 2024. The HateDay dataset, which he coauthored, won an Outstanding Paper award at ACL 2025. Since September 2026, he leads a research group on AI safety and societal impacts at the Hasso Plattner Institute.

Badges

Timeline

  1. 2019–2023

    DPhil at the Oxford Internet Institute

    He starts a DPhil at the Oxford Internet Institute in October 2019, supervised by Helen Margetts and Janet Pierrehumbert. He defends his thesis in May 2023.

  2. 2021

    The HateCheck test suite

    He is first author of HateCheck, a suite of functional tests for hate speech detection models, published at ACL-IJCNLP 2021. The tests cover 29 model functionalities drawn from previous research and interviews with civil society stakeholders.

  3. 2021–2023

    Co-founding Rewire

    He co-founds Rewire in 2021 with Bertie Vidgen, a London startup that builds AI tools for content moderation. ActiveFence acquires the company in March 2023.

  4. June 2023

    Postdoc at Bocconi University

    He joins Dirk Hovy's MilaNLP Lab at Bocconi University as a postdoctoral researcher in June 2023. In January 2024, he publishes SafetyPrompts.com, a catalogue of open datasets for LLM safety.

  5. August 2024

    An ACL Outstanding Paper

    His paper "Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Values and Opinions in Large Language Models", written with Valentin Hofmann, Valentina Pyatkin, Musashi Hinck, Hannah Kirk, Hinrich Schuetze and Dirk Hovy, wins an Outstanding Paper Award at ACL 2024. It challenges multiple-choice evaluations of values and opinions in LLMs and explores more realistic unconstrained evaluations, with the Political Compass Test as a case study.

  6. December 2024

    PRISM, a NeurIPS Best Paper

    The PRISM Alignment Dataset, which he coauthors with Hannah Rose Kirk, Scott A. Hale and others, wins Best Paper in the Datasets and Benchmarks track at NeurIPS 2024. PRISM maps the stated preferences of 1,500 participants from 75 countries to their feedback in 8,011 live conversations with 21 LLMs.

  7. July 2025

    The HateDay dataset

    HateDay, a global hate speech dataset he coauthors with Manuel Tonneau, Scott A. Hale and others, wins an Outstanding Paper award at ACL 2025. It draws on a random sample of all tweets posted on one day in 2022, covers eight languages and four English-speaking countries, and shows that academic datasets greatly overestimate real-world detection performance.

  8. Since September 2026

    A research group at HPI

    He joins the Oxford Internet Institute as Departmental Lecturer in December 2025 and the UK AI Security Institute as a Research Scientist in June 2026. In September 2026, he moves to Berlin to start a research group on AI safety and societal impacts at the Hasso Plattner Institute (HPI), which he leads.

Career

  1. Since 2026University of PotsdamResearch Group Lead
  2. 2026UK AI Security InstituteResearch Scientist
  3. 2025–2026University of OxfordDepartmental Lecturer
  4. From 2023Bocconi UniversityPostdoctoral Researcher
  5. From 2021RewireCo-founder

Education

  1. 2023University of OxfordDPhil

From the biography.

  1. Through 2025, first appointments to public AI advisory and safety bodies peaked in 2023Includes Paul Röttger. We estimate when the people shaping AI entered public advisory and safety bodies, which institutions connect them, and where they worked beforehand. Through 2025, first recorded appointments peaked in 2023 at 42 people.
  2. 11% of AI doctorate holders from 2010 to 2016 founded or led a company or lab by 10 years after graduationIncludes Paul Röttger. We estimate how often AI doctorate holders founded or led a company or lab by 10 years after graduation, how waiting times compare between founders and top jobs, and which universities trained the most.
  3. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes University of Potsdam. We estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.
  4. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes University of Potsdam. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

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

  1. OII | Dr Paul Röttger oii.ox.ac.uk. Checked 22 October 2026.
  2. Paul Röttger's Website paulrottger.com. Checked 22 October 2026.
  3. HateCheck: Functional Tests for Hate Speech Detection Models - ACL Anthology aclanthology.org. Checked 22 October 2026.
  4. Alice (company) - Wikipedia en.wikipedia.org. Checked 22 October 2026.
  5. Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Values and Opinions in Large Language Models - ACL Anthology aclanthology.org. Checked 22 October 2026.
  6. The PRISM Alignment Dataset: What Participatory, Representative and Individualised Human Feedback Reveals About the Subjective and Multicultural Alignment of Large Language Models arxiv.org. Checked 22 October 2026.
  7. NeurIPS 2024 Awards neurips.cc. Checked 22 October 2026.
  8. HateDay: Insights from a Global Hate Speech Dataset Representative of a Day on Twitter - ACL Anthology aclanthology.org. Checked 22 October 2026.
  9. Dr. Paul Röttger | Science Media Center Germany sciencemediacenter.de. Checked 22 October 2026.