The Center for AI Safety (CAIS) is an AI safety nonprofit whose mission is to reduce societal-scale risks from artificial intelligence. It pursues this through three pillars: technical research that identifies dangerous AI behaviors, field-building that supports researchers with computing infrastructure and fellowships, and advocacy that advises policymakers and industry leaders. CAIS is led by executive and research director Dan Hendrycks.
CAIS is widely known for the 2023 Statement on AI Risk, a one-sentence declaration that mitigating the risk of extinction from AI should be a global priority alongside pandemics and nuclear war. The statement was signed by more than 600 AI experts and public figures, including the CEOs of OpenAI, DeepMind, and Anthropic, and CAIS credits it with helping prompt the UK government to host a global AI safety summit in autumn 2023.
The organization also produces concrete evaluation tools. It co-created the WMDP benchmark, which measures hazardous knowledge in language models across biosecurity, cybersecurity, and chemical security, and it has contributed widely used research on robustness, deception, and reliability.
For a general audience, CAIS matters because it sits at the intersection of measurement and public attention, turning technical safety concerns into evidence and statements that policymakers and the public can act on.