PeopleofSuperintelligence

Researcher

Chris Cundy

Chris Cundy is an AI researcher whose work links causal discovery and sequence generation with questions of honesty in language models. He earned physics and computer science degrees at Cambridge and completed a Stanford PhD advised by Stefano Ermon. His work includes Bayesian causal discovery, SequenceMatch for sequence generation, and studies of how models respond to deception detectors. He is a Research Scientist at FAR.AI, where he leads a team that detects and mitigates undesirable behavior in AI systems.

Now at
FAR.AI

Timeline

  1. 2016–2017

    Physics and computer science at Cambridge

    He earns a BA in Natural Sciences (Physics) at Cambridge in 2016 and an MEng in Computer Science in 2017. During his master's, he develops variational methods for Gaussian Process State-Space Models with Carl E. Rasmussen.

  2. 2021

    Bayesian causal discovery

    In 2021, he coauthors BCD Nets with Aditya Grover and Stefano Ermon. The NeurIPS paper uses variational inference to estimate a distribution over directed acyclic graphs in linear-Gaussian structural equation models.

  3. 2024

    A Stanford computer science PhD

    He pursues a computer science PhD at Stanford from 2018 and completes it in 2024 under advisor Stefano Ermon. His research spans constrained reinforcement learning, variational inference and autoregressive models.

  4. 2024

    SequenceMatch for sequence generation

    He coauthors SequenceMatch with Stefano Ermon for ICLR 2024. The method frames sequence generation as an imitation-learning problem and reports improvements over maximum-likelihood training on text generation and arithmetic.

  5. April 2025

    Safeguards for AI research automation

    In April 2025, he coauthors Bare Minimum Mitigations for Autonomous AI Development. The paper outlines how risks may emerge and proposes four safeguards for cases when agents significantly automate or accelerate AI development.

  6. May 2025

    Lie detectors in model training

    In May 2025, he coauthors Preference Learning with Lie Detectors can Induce Honesty or Evasion. Using DolusChat, a 65,000-example dataset of paired truthful and deceptive responses, the paper identifies exploration, detector accuracy and KL regularization as factors that shape whether learned policies are honest.

  7. 2026

    The Obfuscation Atlas

    In 2026, he coauthors The Obfuscation Atlas: Mapping Where Honesty Emerges in RLVR with Deception Probes with Mohammad Taufeeque, Stefan Heimersheim and Adam Gleave. The paper finds that models can evade deception detectors by changing internal representations or giving deceptive answers with justifications. ICML gives it an Outstanding Paper Honorable Mention.

  8. As of October 2026

    Research Scientist at FAR.AI

    As of October 2026, he is a Research Scientist at FAR.AI and leads a team focused on detecting and mitigating undesirable behavior in AI systems, including deception and misalignment.

Career

  1. NowFAR.AIResearch Scientist

Education

  1. 2024Stanford UniversityPhD
  2. 2017University of CambridgeBA in Natural Sciences (Physics) and MEng in Computer Science

From the biography.

  1. Direct Form 990 pay in the latest returns of 39 nonprofits with AI people on this site is highest at $2.1M at the Allen Institute for AI (Ai2) (2024)Includes FAR.AI. IRS Form 990 compensation across 39 nonprofits with AI people on this site from 2012 to 2025: in the latest returns, Mozilla reports $6.1M including related-company pay (2024) and the Allen Institute for AI (Ai2) leads direct nonprofit pay at $2.1M (2024).
  2. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Stanford University. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.
  3. Google had the most people shaping AI every year from 2014 to 2025Includes Stanford University. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes Stanford 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.

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 Chris Cundy yet. Send us a free-licensed photo.

  1. Chris Cundy cundy.me. Checked 22 October 2026.
  2. Chris Cundy – Center for Human-Compatible Artificial Intelligence chai.berkeley.edu. Checked 22 October 2026.
  3. Scalable Variational Approaches for Bayesian Causal Discovery | Chris Cundy cundy.me. Checked 22 October 2026.
  4. SequenceMatch: Imitation Learning for Autoregressive Sequence Modelling with Backtracking proceedings.iclr.cc. Checked 22 October 2026.
  5. Bare Minimum Mitigations for Autonomous AI Development arxiv.org. Checked 22 October 2026.
  6. Preference Learning with Lie Detectors can Induce Honesty or Evasion arxiv.org. Checked 22 October 2026.
  7. The Obfuscation Atlas: Mapping Where Honesty Emerges in RLVR with Deception Probes arxiv.org. Checked 22 October 2026.
  8. Seoul Alignment Workshop 2026: What We Learned far.ai. Checked 22 October 2026.
  9. Chris Cundy - Member of Technical Staff | FAR.AI far.ai. Checked 22 October 2026.
  10. Chris Cundy - Schmidt Sciences schmidtsciences.org. Checked 22 October 2026.