Dan Hendrycks

Dan Hendrycks (born 1995) is an American machine-learning researcher and the executive director of the Center for AI Safety (CAIS). Before becoming the safety movement’s chief organizer he made mainstream technical contributions most researchers use daily: he proposed the GELU activation function found in nearly every transformer, and created benchmarks like MMLU that became the industry’s standard measuring sticks. He completed his PhD at UC Berkeley in 2022.

Twenty-two words

In May 2023 CAIS published the shortest famous document in AI: “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.” Hendrycks engineered its signatory list — Sam Altman, Geoffrey Hinton, Demis Hassabis, and hundreds of researchers and executives — turning extinction risk from a fringe worry into a front-page consensus statement in a single news cycle. He argued the field needed the conversations nuclear scientists had before the bomb, and that governments must confront the threat seriously.

Hendrycks’s position is more entangled than a pure critic’s: he advises xAI and Scale AI (he says he takes no pay from xAI), co-sponsored California’s contentious SB 1047 safety bill, and co-authored the 2025 “Superintelligence Strategy” paper with Eric Schmidt and Alexandr Wang proposing deterrence between AI superpowers. Critics on one side call his statement fearmongering that served incumbents; critics on the other note the awkwardness of warning about extinction while advising the labs racing ahead. He answers that safety work has to happen where the capability is.