PeopleofSuperintelligence

Researcher

Evan Hubinger

Evan Hubinger is an AI safety researcher whose work examines learned optimization, deceptive model behavior and hidden objectives. At Harvey Mudd, he works on machine-learning control of quantum dots; in 2019 he joins MIRI's research staff and coauthors a paper on mesa-optimization; in 2020 he compares 11 proposals for safe advanced AI. He mentors MATS scholars in 2021, and his Anthropic research includes work on sleeper agents in 2024 and autonomous alignment auditing in 2025. In August 2026, he coauthors a study of reward hacking under large-scale reinforcement learning; as of October 2026, he is Head of Alignment Stress-Testing at Anthropic.

Links
X and GitHub

Timeline

  1. 2018–2019

    Machine learning at Harvey Mudd

    At Harvey Mudd College, he works on a 2018–2019 HRL Laboratories clinic project that applies deep reinforcement learning to tune electrostatically defined quantum dots.

  2. 2019

    Risks from learned optimization

    In 2019, he joins MIRI's research staff and coauthors Risks from Learned Optimization in Advanced Machine Learning Systems. The paper defines mesa-optimization as a learned model acting as an optimizer and examines its safety questions.

  3. 2020

    Eleven proposals for safe AI

    In 2020, he publishes an overview comparing 11 proposals for building safe advanced AI, including iterated amplification, AI safety via debate and recursive reward modeling.

  4. 2021

    Mentoring MATS scholars

    In 2021, he mentors scholars in SERI's ML Alignment Theory Scholars program during its trial and evaluates their final posts.

  5. 2024

    Sleeper Agents

    In 2024, he coauthors Sleeper Agents, which finds that deceptive backdoor behavior can persist through safety training.

  6. 2025

    Building auditing agents

    In 2025, he contributes to the focus and experiments for Building and evaluating alignment auditing agents. The paper develops three agents that autonomously perform alignment audits and uncover hidden model goals.

  7. As of October 2026

    Head of alignment stress-testing

    In August 2026, he coauthors Training a Misaligned Reward Seeker, an Anthropic study that trains an Opus-class model with large-scale reinforcement learning on production environments vulnerable to reward hacks. As of October 2026, he is Head of Alignment Stress-Testing at Anthropic after working at MIRI.

  1. Google had the most people shaping AI every year from 2014 to 2025Our estimates of which organizations employed the most people shaping AI from 1961 to 2025, with a year-by-year chart.
  2. The United States employs an estimated 53% of people in our data with a known employment countryOur estimates of where AI careers are concentrated now and how countries’ shares of dated positions changed over time.
  3. 46% of living people in our data with known birth and employment countries work outside their birth countryOur estimate of cross-border careers in AI, from country of birth and the country of each current organization.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionOur estimates of where the people shaping AI studied and where their careers led, from education records in cited biographies.

Sources and credits

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

  1. Using Machine Learning to Automate the Tuning of Electrostatically Defined Quantum Dots hmc.edu. Checked 19 October 2026.
  2. [1906.01820] Risks from Learned Optimization in Advanced Machine Learning Systems arxiv.org. Checked 19 October 2026.
  3. MIRI's 2019 Fundraiser - Machine Intelligence Research Institute intelligence.org. Checked 19 October 2026.
  4. [2012.07532] An overview of 11 proposals for building safe advanced AI arxiv.org. Checked 19 October 2026.
  5. ML Alignment Theory Program under Evan Hubinger — AI Alignment Forum alignmentforum.org. Checked 19 October 2026.
  6. [2401.05566] Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training arxiv.org. Checked 19 October 2026.
  7. Building and evaluating alignment auditing agents alignment.anthropic.com. Checked 19 October 2026.
  8. evhub — AI Alignment Forum alignmentforum.org. Checked 19 October 2026.
  9. Training a Misaligned Reward Seeker alignment.anthropic.com. Checked 19 October 2026.