PeopleofSuperintelligence

Researcher

Katherine Lee

Katherine Lee is an AI researcher whose work connects language-model development with questions of memorization, security, privacy and copyright. At Princeton, she studied operations research and explored neural networks with Jonathan Pillow and Uri Hasson. She joined Google Brain in 2017, helped develop Google's T5, and examined how models reproduce their training data. Her 2024 paper “Stealing Part of a Production Language Model” received an ICML Best Paper award, and she coauthored a 2025 NeurIPS analysis of machine unlearning. As of October 2026, she is a researcher at OpenAI and runs the GenLaw Center.

Now at
OpenAI and GenLaw Center
Works
52
Badges

Publication numbers from OpenAlex, as of 22 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

  1. 2017

    Joining Google Brain

    During her undergraduate study in operations research at Princeton, she explores neural networks with Jonathan Pillow and Uri Hasson. She joins Google Brain in 2017.

  2. 2020

    T5 and text-to-text learning

    In 2020, she coauthors Google Research’s T5 model and helps introduce the C4 pretraining dataset.

  3. 2021

    Extracting GPT-2 training data

    In 2021, she coauthors a USENIX Security paper showing that queries to GPT-2 can recover hundreds of verbatim training sequences.

  4. 2022

    Deduplicating training data

    In 2022, she coauthors an ACL study that develops two tools to deduplicate training datasets.

  5. 2023

    Google DeepMind and AI copyright

    In 2023, she moves to Google DeepMind and coauthors “Talkin’ ’Bout AI Generation,” which examines the copyright consequences of design choices across generative AI’s stages.

  6. 2024

    An ICML Best Paper

    In 2024, she coauthors “Stealing Part of a Production Language Model,” which introduces an attack that extracts precise information from black-box language models. The paper receives an ICML Best Paper award.

  7. 2025

    Machine unlearning at NeurIPS

    In 2025, she works at Google DeepMind and coauthors a NeurIPS analysis that finds a gap between machine-unlearning methods and the legal and policy goals people ask them to meet.

  8. As of October 2026

    OpenAI and GenLaw

    As of October 2026, she is a researcher at OpenAI and runs the GenLaw Center. She studies security and privacy in generative AI. In 2026, she coauthors a method that uses decoding-constrained beam search to estimate near-verbatim extraction risk at lower cost.

Career

  1. NowOpenAIResearcher
  2. NowGenLaw Center
  3. From 2023Google DeepMind
  4. From 2017Google Brain

Education

  1. Princeton University

From the biography.

  1. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Katherine Lee. 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.
  2. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Katherine Lee. 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.
  3. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Katherine Lee. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  4. 16% of AI career moves in the 2020s went from a university to a companyIncludes Katherine Lee. 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 Katherine Lee yet. Send us a free-licensed photo.

  1. Katherine Lee katelee168.github.io. Checked 20 October 2026.
  2. Exploring Transfer Learning with T5: the Text-To-Text Transfer Transformer research.google. Checked 20 October 2026.
  3. Extracting Training Data from Large Language Models usenix.org. Checked 20 October 2026.
  4. Deduplicating Training Data Makes Language Models Better - ACL Anthology aclanthology.org. Checked 20 October 2026.
  5. Talkin' 'Bout AI Generation: Copyright and the Generative-AI Supply Chain arxiv.org. Checked 20 October 2026.
  6. Stealing Part of a Production Language Model - Microsoft Research microsoft.com. Checked 20 October 2026.
  7. Machine Unlearning Doesn’t Do What You Think: Lessons for Generative AI Policy, Research, and Practice - Microsoft Research microsoft.com. Checked 20 October 2026.
  8. The Impact of Memorization on Trustworthy Foundation Models icml2026memfm.github.io. Checked 20 October 2026.
  9. Estimating near-verbatim extraction risk in language models with decoding-constrained beam search arxiv.org. Checked 20 October 2026.