METR (Model Evaluation and Threat Research) is a nonprofit research institute that scientifically measures whether and when advanced AI systems might pose catastrophic societal harm, partnering with leading AI companies like OpenAI and Anthropic for pre-deployment risk assessments.