Paul Christiano is an American AI safety researcher. He received a PhD in theoretical computer science from the University of California, Berkeley, and led the language model alignment team at OpenAI. In 2021 he founded the Alignment Research Center (ARC), a nonprofit that evaluates risks from AI systems, and in 2024 he became head of AI safety at the US AI Safety Institute within the Department of Commerce.
With Jan Leike, Dario Amodei, and others he co-authored the 2017 paper "Deep Reinforcement Learning from Human Preferences," which laid the groundwork for RLHF by using human preference comparisons as a reward signal. He has also proposed scalable oversight methods such as iterated amplification.