METR

METR (Model Evaluation and Threat Research) is a nonprofit that evaluates frontier AI systems to help developers and policymakers understand autonomous capabilities and catastrophic risks. Its assessments test whether agents can conduct research, develop applications, carry out cyberattacks or resist shutdown. Its Inspect Hawk platform lets evaluation teams run Inspect AI tasks on AWS, managing isolated execution, credentials, logs and results with a web interface. Hawk supplies operational infrastructure around the UK AI Safety Institute’s evaluation framework.

Founded by Beth Barnes in 2022 as ARC Evals within the Alignment Research Center, the organization adopted the METR name in 2023; Barnes is its current CEO. Its task-completion time horizons measure the human-expert task duration at which an agent is predicted to succeed with a specified probability. Based primarily on software engineering, machine-learning and cybersecurity tasks, these measurements describe task difficulty, rather than uninterrupted agent runtime.

METR helped prototype the Responsible Scaling Policies approach. In August 2026, it reported approximately $71 million in funding commitments raised over the preceding six months to expand its team and research. METR is donation-funded, with a small European AI Office technical-assistance contract. It has not accepted funding from AI companies, although they provide significant free tokens for evaluations, research and engineering.