Top AI safety organizations

METR measures what frontier AI agents can do in realistic, extended tasks rather than short benchmark questions. Its widely followed time-horizon evaluations estimate the length of software and research tasks models can complete, giving labs and policymakers a concrete view of rapidly changing autonomous capability.

San Francisco non-profit focused on reducing societal-scale risks from AI through research and field-building.

UK government body (formerly the AI Safety Institute) that evaluates risks from advanced AI models.

Apollo Research tests whether advanced models can deceive evaluators, pursue hidden goals or evade human control. Its behavioral evaluations and interpretability research provide frontier labs and policymakers with concrete evidence about forms of model behavior that ordinary benchmarks miss.

Alignment Research Center develops methods for keeping powerful AI systems helpful and honest as their capabilities grow. Its work helped define scalable alignment and eliciting latent knowledge—ways to test whether models know more than they reveal—and its former evaluations team became METR.

AI alignment research organization focused on empirical safety work, interpretability, adversarial robustness, and failure modes in advanced models.

AI safety research nonprofit founded by Adam Gleave and Karl Berzins, based in Berkeley. Works on adversarial robustness, model evaluation, interpretability, and alignment, and convenes the AI safety research community.

Research organization producing data and analysis on AI trends, compute, benchmarks, timelines, and the trajectory of machine-learning capabilities.

Independent research organization producing policy research and developing talent to help governments, technology companies, and other institutions navigate the risks and opportunities of advanced AI.

Berkeley-based AI safety research organization focused on theoretical alignment, decision theory, and risks from advanced machine intelligence.

Non-profit working to reduce catastrophic and existential risks from transformative technologies, especially AI. Known for the Asilomar AI Principles and the 2023 open letter calling for a pause on giant AI experiments.

AI governance and safety organisation co-founded by Yoshua Bengio. Works to develop international frameworks for AI oversight and advocates for responsible AI development policy, with a focus on existential and catastrophic risk.