Center for AI Safety conducts safety research, develops evaluation benchmarks, and supports researchers with compute and education to reduce societal-scale risks from AI.