Jan Leike is a machine learning and alignment researcher at Anthropic, best known for his work on aligning large language models with human intent. Until early 2026 he led Anthropic's Alignment team, working on scalable oversight, character training, Constitutional Classifiers and improving Claude's alignment. He is now working on a new project at Anthropic. Before that he co-led the Superalignment team at OpenAI, where he helped develop InstructGPT, ChatGPT and the alignment of GPT-4, and co-authored the team's research roadmap. At DeepMind he prototyped reinforcement learning from human feedback. He holds a PhD in reinforcement learning theory from the Australian National University, and TIME named him one of the 100 most influential people in AI in 2023 and 2024.
Alignment Researcher (formerly Alignment team lead)
Led Anthropic's Alignment team until early 2026 (scalable oversight, character training, Constitutional Classifiers, Claude's alignment); now working on a new project at Anthropic.
Head of Alignment; Superalignment Co-lead
Led alignment research; co-led the Superalignment team from 2023. Contributed to InstructGPT, ChatGPT, and GPT-4 alignment.
Alignment Researcher
Researched AI alignment and prototyped reinforcement learning from human feedback.
Postdoctoral Researcher
Six-month postdoc after his PhD.
PhD, Computer Science (reinforcement learning theory; advisor Marcus Hutter)
BSc and MSc
