PeopleofSuperintelligence

Researcher

Hannah Rose Kirk

Also known as Hannah Kirk

Hannah Rose Kirk is an AI researcher who studies how human feedback aligns language models and how frontier AI shapes the people who use it. After degrees at Cambridge and Peking University, she began a DPhil in Social Data Science at the Oxford Internet Institute in 2021, worked in online safety at The Alan Turing Institute, and was first author of the Hatemoji dataset paper in 2022. In 2024 she was first author of the PRISM Alignment Dataset paper, built from 1,500 participants in 75 countries, which won the Best Paper Award in the NeurIPS 2024 Datasets & Benchmarks track. Since September 2024, she is Workstream Lead for Human Influence at the UK AI Security Institute, investigating the social and psychological capabilities of frontier AI.

Timeline

  1. 2018–2020

    Cambridge and Peking University degrees

    Graduating from Trinity College, University of Cambridge in 2018 with a double first in Economics, she moves to Beijing for two years. There she earns an MA at Yenching Academy, Peking University in 2020 with honours and works at a think-tank on eastern and western philosophical approaches to artificial intelligence.

  2. 2020–2021

    Social Data Science at Oxford

    From October 2020 to July 2021 she studies for an MSc in Social Data Science at the Oxford Internet Institute and finishes with Distinction. In September 2021 she begins a DPhil in Social Data Science there, supervised by Scott A. Hale and Bertie Vidgen.

  3. 2021–2023

    Online safety at the Turing Institute

    From September 2021 to September 2023 she is a data scientist in online safety at The Alan Turing Institute, where she builds data-centric AI to detect online harms efficiently.

  4. 2022

    The Hatemoji dataset

    She is first author of Hatemoji, a test suite and adversarially-generated dataset for benchmarking and detecting emoji-based hate, published at NAACL in 2022.

  5. 2023

    Safety work at OpenAI and Google

    In 2023, she is an external student researcher at Google, where she co-hosts an adversarial challenge to identify unsafe failure modes in text2image models. From August 2023 to December 2023 she is also a red-teamer and consultant at OpenAI, improving the safety of DALL-E and GPT-4, and from September 2023 to December 2023 a visiting academic in data science at New York University, collaborating on human-AI coordination and LLM alignment.

  6. April 2024

    A Nature Machine Intelligence perspective

    In April 2024, Nature Machine Intelligence publishes her perspective, written with Bertie Vidgen, Paul Röttger and Scott A. Hale, on the benefits, risks and bounds of personalizing the alignment of large language models to individuals. It sets out a taxonomy of risks and benefits of personalized LLMs for individuals and society.

  7. 2024

    PRISM and a NeurIPS best paper

    With supervisors Scott A. Hale and Bertie Vidgen, she launches the PRISM alignment project, and in 2024 she is first author of the PRISM Alignment Dataset paper, which maps the stated preferences of 1,500 participants from 75 countries to their feedback in 8,011 live conversations with 21 LLMs. The paper receives the Best Paper Award in the Datasets & Benchmarks track at NeurIPS 2024. Co-authored work on political bias in AI systems also earns her an Outstanding Paper Award at ACL 2024.

  8. Since September 2024

    Human Influence at the AI Security Institute

    She joins the UK AI Security Institute in September 2024 and is Workstream Lead for Human Influence, investigating the social and psychological capabilities of frontier AI. In December 2025 she is first author of a study with Christopher Summerfield and others that combines longitudinal randomised controlled trials (N=3,534) with neural steering vectors to vary exposure to relationship-seeking AI. In June 2026 she is among the authors of "AI systems out-persuade expert humans", whose four preregistered experiments find AI more persuasive than expert humans.

Career

  1. Since 2024UK AI Security InstituteWorkstream Lead for Human Influence
  2. 2023GoogleExternal Student Researcher
  3. 2023OpenAIRed-teamer and Consultant
  4. 2023New York UniversityVisiting Academic
  5. 2021–2023Alan Turing InstituteData Scientist, Online Safety

Education

  1. 2021University of OxfordMSc in Social Data Science
  2. From 2021University of OxfordDPhil in Social Data Science
  3. 2020Peking UniversityMA
  4. 2018Trinity College, University of Cambridge

From the biography.

  1. Through 2025, first appointments to public AI advisory and safety bodies peaked in 2023Includes Hannah Rose Kirk. 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. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Hannah Rose Kirk. Our estimates of where OpenAI recruits came from and where 133 people with former OpenAI roles work now, including the companies and labs they founded.
  3. Google had the most people shaping AI every year from 2014 to 2025Includes Google. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  4. Google has the highest combined h-index among AI organizationsIncludes Google. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.

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 Hannah Rose Kirk yet. Send us a free-licensed photo.

  1. DemTech | Hannah Kirk demtech.oii.ox.ac.uk. Checked 21 October 2026.
  2. Hannah Rose Kirk - Oxford Sparks oxfordsparks.ox.ac.uk. Checked 21 October 2026.
  3. Hannah Rose Kirk hannahrosekirk.com. Checked 21 October 2026.
  4. OII | Hannah Rose Kirk oii.ox.ac.uk. Checked 21 October 2026.
  5. The benefits, risks and bounds of personalizing the alignment of large language models to individuals | Nature Machine Intelligence nature.com. Checked 21 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 21 October 2026.
  7. Professor He He and PhD Candidate Hannah Rose Kirk Win NeurIPS 2024 Best Paper Award for PRISM Alignment Dataset cims.nyu.edu. Checked 21 October 2026.
  8. How we can better align Large Language Models with diverse humans oii.ox.ac.uk. Checked 21 October 2026.
  9. Neural steering vectors reveal dose and exposure-dependent impacts of human-AI relationships arxiv.org. Checked 21 October 2026.
  10. AI systems out-persuade expert humans arxiv.org. Checked 21 October 2026.