Who's who · AI security & safety
METR
Non-profit that measures how well frontier AI models can carry out long tasks on their own
At a glance
- Founded
- 2022
- Based in
- Berkeley, United States
- People
- Beth Barnes, Founder and chief executive; Chris Painter, President; Hjalmar Wijk, Chief scientist; Nate Rush, Chief technology officer
- Known for
- Time horizon measurements, Pre-deployment evaluations, Developer productivity study
- Funding
- Non-profit; about $71 million in commitments raised in six months (Aug 2026); no frontier AI company money
- Former name
- ARC Evals (until December 2023)
- Website
- https://metr.org
METR (Model Evaluation and Threat Research) is a non-profit that evaluates frontier AI models to inform the public about their risks and capabilities. It develops scientific methods to assess catastrophic risks from autonomous capabilities, meaning what an AI system can do without step-by-step human direction.
Beth Barnes founded the work in 2022 as ARC Evals, a team under the Alignment Research Center, and it became an independent organisation called METR in December 2023. It is best known for its "time horizon" measure, the length of task an AI agent can complete, which it reports has doubled roughly every seven months for six years. It has also measured how AI tools affect the speed of experienced software developers.
METR does not accept money from frontier AI companies or donations directed by their staff, though those companies give it free model access. In August 2026 it reported about $71 million in commitments over the previous six months, from The Audacious Project, foundations and individual donors. Its partners include OpenAI, Anthropic, Google DeepMind and the European AI Office.
Sources
- METR, About METR, 2026-10-02
- METR, Funding update, 2026-08-14
- METR, Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity, 2025-07-10
- TIME, Nobody Knows How to Safety-Test AI, 2024-03-21
Facts last checked 2026-10-02