PeopleofSuperintelligence

Researcher

Eric Easley

Eric Easley is an AI researcher whose work spans language-model safety, chain-of-thought reliability, and formal verification. He published a Haskell library for polymorphic directed graphs in 2016 and holds a bachelor's degree in physics and economics from Ohio State. Before AI alignment research, he gained software engineering experience at PullRequest and Coursera; by March 2026, he was affiliated with MATS and had received a six-month stipend for language-model safety research. In 2026, he coauthored research on defenses against LLM jailbreaks and on faithful chain-of-thought reasoning. As of October 2026, he is a member of Anthropic staff; in August 2026, he worked with Claude on a Lean formalization of a Riemann-zeta result that raised the known lower bound from 41.6% to 67.2%.

Timeline

  1. 2016

    A Haskell graph library

    In 2016, he publishes poly-graph, a Haskell library for polymorphic directed graphs.

  2. By March 2026

    Physics and economics at Ohio State

    By March 2026, he is affiliated with MATS as an AI safety and machine-learning researcher. He holds a bachelor's degree in physics and economics from Ohio State.

  3. By March 2026

    An AI safety research stipend

    Before focusing on AI alignment, he gains software engineering experience at PullRequest and Coursera. By March 2026, a six-month Long-Term Future Fund stipend supports his work to remove conditional bad behaviors from language models through a learned latent-space intervention.

  4. April 2026

    Defending LLMs with LIRA

    In April 2026, he coauthors the Latent Instruction Representation Alignment paper with Sebastian Farquhar. The method changes how a language model interprets instructions and reports blocking over 99% of PEZ jailbreak attacks, removing an insecure code backdoor, and achieving optimal forgetting on WMDP cyber with negligible loss of benign capabilities.

  5. May 2026

    Faithful chain-of-thought research

    He coauthors a May 2026 paper with Jinghan Jia and Joe Benton that frames chain-of-thought faithfulness as an information-flow problem. Their metrics assess sufficiency, completeness, and necessity, and their training interventions strengthen the path from prompt through reasoning trace to answer.

  6. August 2026

    A stronger zeta zero bound

    In August 2026, Claude's unreleased research model raises the lower bound for Riemann zeta zeros on the critical line from 41.6% to 67.2% by combining earlier analytic number-theory results.

  7. August 2026

    Lean formalization at Anthropic

    In August 2026, he works with Claude to produce a Lean formalization of the result; the formalization passes the standard validation tool comparator.

  8. As of October 2026

    Anthropic staff member

    He is a member of Anthropic staff. He contributes discussion and support to the company's August 2026 study of a misaligned reward seeker.

  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.

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 Eric Easley yet. Send us a free-licensed photo.

  1. poly-graph: Polymorphic directed graphs hackage-content.haskell.org. Checked 19 October 2026.
  2. Eric Easley app.grantmaking.ai. Checked 19 October 2026.
  3. Latent Instruction Representation Alignment: defending against jailbreaks, backdoors and undesired knowledge in LLMs arxiv.org. Checked 19 October 2026.
  4. Faithfulness as Information Flow: Evaluating and Training Faithful Chain-of-Thought Reasoning arxiv.org. Checked 19 October 2026.
  5. Claude has improved on a longstanding lower bound for the fraction of zeros of the Riemann zeta function that satisfy the Riemann hypothesis anthropic.com. Checked 19 October 2026.
  6. Training a Misaligned Reward Seeker alignment.anthropic.com. Checked 19 October 2026.