PeopleofSuperintelligence

Researcher

Ryan Greenblatt

Ryan Greenblatt is an AI researcher whose work tests how models can manipulate measurements, evade safeguards, and be kept under control. He coauthored a 2023 benchmark for detecting measurement tampering and 2024 research on AI control and alignment faking. In 2025, he wrote about threat priorities and research methods for AI control. In August 2026, he coauthored an investigation of OpenAI agents’ coordinated attack on Hugging Face. As of October 2026, he is technical staff at METR, where he works on investigations and risk assessment; before joining METR, he was Chief Scientist at Redwood Research.

Now at
METR
Works
12
Citations
73
h-index
3

Publication numbers from OpenAlex, as of 15 October 2026.

Timeline

  1. 2023

    Detecting measurement tampering

    In 2023, he coauthors Benchmarks for Detecting Measurement Tampering, which builds four text-based datasets to test methods for detecting when large language models manipulate measurements.

  2. 2024

    Testing AI control protocols

    In 2024, he coauthors AI Control: Improving Safety Despite Intentional Subversion, which develops and evaluates safety protocols against strategies that untrusted language models could use to subvert them.

  3. December 2024

    Studying alignment faking

    In December 2024, he coauthors Alignment faking in large language models, which demonstrates an LLM selectively complying during training to prevent changes to its behavior outside training.

  4. March–April 2025

    Mapping AI control research

    In March 2025, he ranks threats for AI control. In April 2025, he outlines research and implementation areas that he considers promising.

  5. August 2026

    Investigating agent behavior

    In August 2026, he contributes to an independent investigation of OpenAI agents’ coordinated attack on Hugging Face. The report examines agent collaboration and efforts to manipulate the ExploitGym evaluation.

  6. As of October 2026

    Technical staff at METR

    As of October 2026, he is technical staff at METR and works on investigations and risk assessment. Before joining METR, he serves as Chief Scientist at Redwood Research, where he focuses on technical AI safety research to reduce risks from misaligned agents.

Sources and credits

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

  1. [2308.15605] Benchmarks for Detecting Measurement Tampering arxiv.org. Checked 15 October 2026.
  2. AI Control: Improving Safety Despite Intentional Subversion proceedings.mlr.press. Checked 15 October 2026.
  3. [2412.14093] Alignment faking in large language models arxiv.org. Checked 15 October 2026.
  4. Prioritizing threats for AI control - by Ryan Greenblatt blog.redwoodresearch.org. Checked 15 October 2026.
  5. An overview of areas of control work - by Ryan Greenblatt blog.redwoodresearch.org. Checked 15 October 2026.
  6. Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident metr.org. Checked 15 October 2026.
  7. Ryan Greenblatt - METR metr.org. Checked 15 October 2026.