PeopleofSuperintelligence

Researcher

Katarina Slama

Katarina Slama is an AI researcher whose work connects human attention research, language-model alignment and the societal impacts of AI. Her path ran from small-town Finland to US universities, where she earned a UC Berkeley neuroscience PhD focused on human attention; in 2021, she led a study of visual search using intracranial recordings. At OpenAI, she contributed to InstructGPT, published in 2022, and coauthored OpenAI's GPT-4 Technical Report in 2023. Her later research compared AI-scheming research with 1970s studies of ape language and tested how language-model preferences shape donation advice and refusals. As of October 2026, she is an AI Societal Impacts Research Scientist at the UK AI Security Institute; in September 2026, she signed a statement calling for limits on recursive self-improvement until it can be done safely.

Badges

Timeline

  1. 2021

    Visual search at Berkeley

    Her path runs from small-town Finland to US universities, and she pivots from neuroscience research to AI. In 2021, she leads a visual-search study in Robert Knight's lab at UC Berkeley's Helen Wills Neuroscience Institute, using intracranial recordings from patients awaiting surgery for epilepsy.

  2. 2022

    InstructGPT with human feedback

    At OpenAI, she contributes to the development of InstructGPT, published in 2022. The paper fine-tunes GPT-3 with labeler demonstrations and rankings of model outputs, then uses reinforcement learning from human feedback.

  3. March 2023

    The GPT-4 technical report

    In March 2023, she coauthors OpenAI's GPT-4 Technical Report, which describes GPT-4 as a multimodal model that accepts image and text inputs and produces text outputs.

  4. July 2025

    Comparing AI scheming research

    In July 2025, she coauthors a paper comparing research on AI scheming with 1970s tests of whether non-human primates could master natural language. It argues that research into AI scheming should avoid overattributing human traits to other agents.

  5. August 2026

    Testing LLM preferences

    In August 2026, she coauthors a TMLR paper on whether language-model preferences predict downstream behavior. Across five frontier LLMs, the study finds that all five give preference-aligned donation advice and refuse more often for less-preferred entities.

  6. September 2026

    Limits on recursive self-improvement

    On 24 September 2026, she signs a civil-society statement urging governments to limit recursive self-improvement until it can be done safely.

  7. As of October 2026

    AI societal impacts at UK AISI

    In December 2025, she researches societal impacts of transformative AI with the UK AI Security Institute and holds a UC Berkeley neuroscience PhD focused on mechanisms of human attention allocation. As of October 2026, she is an AI Societal Impacts Research Scientist at the institute.

Career

  1. Since 2025UK AI Security InstituteAI Societal Impacts Research Scientist
  2. 2022OpenAI
  3. 2021University of California, Berkeley

Education

  1. University of California, BerkeleyPhD in Neuroscience

From the biography.

  1. Through 2025, first appointments to public AI advisory and safety bodies peaked in 2023Includes Katarina Slama. 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 Katarina Slama. 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. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Katarina Slama. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  4. 16% of AI career moves in the 2020s went from a university to a companyIncludes Katarina Slama. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

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

  1. Podcast about artificial intelligence: Ja ihminen loi älyn | Aalto University aalto.fi. Checked 21 October 2026.
  2. Looking for something? Recording directly from the human brain reveals mechanisms underlying visual search | Helen Wills Neuroscience Institute hwni.berkeley.edu. Checked 21 October 2026.
  3. Training language models to follow instructions with human feedback proceedings.neurips.cc. Checked 21 October 2026.
  4. Introduction to AI Safety Institutes with Tutke and researchers from UK AISI luma.com. Checked 21 October 2026.
  5. GPT-4 Technical Report arxiv.org. Checked 21 October 2026.
  6. Lessons from a Chimp: AI "Scheming" and the Quest for Ape Language arxiv.org. Checked 21 October 2026.
  7. Lessons from a chimp: AI "scheming" and the quest for ape language aisi.gov.uk. Checked 21 October 2026.
  8. Transactions on Machine Learning Research jmlr.org. Checked 21 October 2026.
  9. When Do LLM Preferences Predict Downstream Behavior? aisi.gov.uk. Checked 21 October 2026.
  10. Statement on RSI 2026 — AI Safety Connect viola-reed-e2ka.squarespace.com. Checked 21 October 2026.