PeopleofSuperintelligence

Researcher

Peter Henderson

Peter Henderson is a computer scientist with a law degree whose research connects reinforcement learning, AI safety and AI law. After working as a software engineer on Alexa and AWS, he studied at McGill University and Mila and wrote "Deep Reinforcement Learning that Matters," published at AAAI in 2018. He earned a J.D. and a Ph.D. in computer science at Stanford, released the Pile of Law legal dataset in 2022, and joined Princeton in 2024. A 2024 paper on how fine-tuning compromises the safety of aligned language models was an ICLR oral, and a 2025 paper on shallow safety alignment won an ICLR Outstanding Paper award. He is an assistant professor at Princeton, where he leads the Polaris Lab.

Timeline

  1. By 2018

    Deep reinforcement learning that matters

    He works as a software engineer on Alexa and AWS, then earns a master's degree at McGill University and Mila, where he does reinforcement learning research with David Meger and Joelle Pineau. With Riashat Islam, Philip Bachman, Pineau, Doina Precup and Meger, he writes "Deep Reinforcement Learning that Matters," posted on arXiv in 2017 and published at AAAI in 2018. The paper examines how hard deep RL results are to reproduce and suggests guidelines for more reproducible reporting.

  2. 2020

    Reporting machine learning's carbon footprint

    In 2020, the Journal of Machine Learning Research publishes "Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning," which he writes with Jieru Hu, Joshua Romoff, Emma Brunskill, Dan Jurafsky and Joelle Pineau. The paper introduces a framework for tracking real-time energy use and carbon emissions of experiments and a leaderboard for energy-efficient reinforcement learning algorithms.

  3. 2022

    The Pile of Law dataset

    In 2022, he posts "Pile of Law" with Mark Krass, Lucia Zheng, Neel Guha, Christopher Manning, Dan Jurafsky and Daniel Ho. The paper releases a 256GB open-source dataset of court opinions, contracts, administrative rules and legislative records, and presents an approach to filtering pretraining data that is grounded in law.

  4. 2024

    Assistant professor at Princeton

    He joins Princeton in 2024 as an assistant professor with joint appointments in the Department of Computer Science and the School of Public and International Affairs, and a place at the Center for Information Technology Policy. He comes from Stanford, where he earns a J.D. from Stanford Law School and a Ph.D. in computer science, working with Dan Jurafsky.

  5. 2024

    ICLR oral on fine-tuning risks

    His paper on fine-tuning aligned language models, written with Xiangyu Qi, Yi Zeng, Tinghao Xie, Pin-Yu Chen, Ruoxi Jia and Prateek Mittal, is an oral at ICLR 2024. Its red-teaming studies jailbreak GPT-3.5 Turbo's safety guardrails with only 10 adversarially designed examples at a cost of less than $0.20, and show that fine-tuning on benign datasets can also degrade safety alignment.

  6. 2025

    An ICLR Outstanding Paper

    In 2025, his paper on the depth of safety alignment, written with Xiangyu Qi, Ashwinee Panda, Kaifeng Lyu, Xiao Ma, Subhrajit Roy, Ahmad Beirami and Prateek Mittal, is one of three ICLR 2025 Outstanding Papers. It argues that safety alignment can adapt a model's output over only its first few tokens, calls this shallow safety alignment, and proposes a regularized fine-tuning objective that makes alignment more persistent against fine-tuning attacks.

  7. July 2026

    A library for public defenders

    In July 2026, the New Jersey Office of the Public Defender launches the OPD Resource Library, an AI-powered search platform for briefs, motions and legal research that Henderson and Dominik Stammbach engineer. The tool searches an internal library of 3,000 briefs, 200 internal documents and 350 directives, and shows the original PDF with the relevant section highlighted.

  8. As of October 2026

    Leading the Polaris Lab

    He is an assistant professor of computer science and of the Princeton School of Public and International Affairs. He leads the Polaris Lab and works on reinforcement learning, alignment and law.

Career

  1. Since 2024Princeton UniversityAssistant Professor
  2. MilaResearcher

Education

  1. McGill UniversityMaster's degree
  2. Stanford UniversityJ.D.

From the biography.

  1. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Princeton University. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  2. 84% of people shaping AI with recorded undergraduate and doctoral schools changed institutions for their doctorateIncludes Princeton University. 84% of people shaping AI with recorded undergraduate and doctoral schools changed institutions for their doctorate. Across 2,116 people with both an undergraduate degree and a doctorate recorded, we trace which undergraduate schools feed which doctoral programs, how often students stay at the same school, and how paths differ by birth-country region.
  3. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes Princeton University. 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 median h-index is 48 at universities and 14 at frontier AI labsIncludes Princeton University. 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.

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 Peter Henderson yet. Send us a free-licensed photo.

  1. AI and Law Expert to Join SPIA from Stanford in January spia.princeton.edu. Checked 22 October 2026.
  2. Peter Henderson peterhenderson.co. Checked 22 October 2026.
  3. [1709.06560v3] Deep Reinforcement Learning that Matters arxiv.org. Checked 22 October 2026.
  4. Deep Reinforcement Learning That Matters ojs.aaai.org. Checked 22 October 2026.
  5. Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning jmlr.org. Checked 22 October 2026.
  6. [2207.00220] Pile of Law: Learning Responsible Data Filtering from the Law and a 256GB Open-Source Legal Dataset arxiv.org. Checked 22 October 2026.
  7. [2310.03693] Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To! arxiv.org. Checked 22 October 2026.
  8. ICLR Oral Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To! iclr.cc. Checked 22 October 2026.
  9. Announcing the Outstanding Paper Awards at ICLR 2025 blog.iclr.cc. Checked 22 October 2026.
  10. [2406.05946] Safety Alignment Should Be Made More Than Just a Few Tokens Deep arxiv.org. Checked 22 October 2026.
  11. New Jersey Office of the Public Defender Launches OPD Resource Library nj.gov. Checked 22 October 2026.
  12. Peter Henderson - Princeton Engineering engineering.princeton.edu. Checked 22 October 2026.