Jan Leike

Career

1. 
   Anthropic

   • Alignment Science team leadMay 2024 – present
2. 
   OpenAI

   • Head of Alignment; co-led the Superalignment team2021–2024
3. 
   DeepMind

   • Alignment researcherbefore 2021
4. 
   Future of Humanity Institute, University of Oxford

   • Postdoctoral fellowship, six months

Education

• BS, University of Freiburg
• MS, computer science, University of Freiburg
• PhD, machine learning, Australian National University

Lesser-known facts

• Did a PhD at the Australian National University on AIXI, a mathematical model of general intelligence.
• Named to TIME's 100 most influential people in AI in 2023 and 2024.
• Worked on InstructGPT, ChatGPT and the alignment of GPT-4 at OpenAI, and wrote its approach to alignment research.
• Superalignment, which he co-led with Ilya Sutskever, aimed to align superintelligence within four years using 20% of OpenAI's compute.
• Joined Anthropic on May 28, 2024, two weeks after resigning from OpenAI.
• Announced in April 2026 that Claude made fully autonomous progress on scalable oversight research, outperforming human researchers for $18,000 in compute.

Research papers before Anthropic

• Deep Reinforcement Learning from Human Preferences2017 · author 2 of 6

Known for

• Deep Reinforcement Learning from Human Preferences (2017)
• Scalable agent alignment via reward modeling: a research direction (2018)
• Training language models to follow instructions with human feedback (2022)
• Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision (2024)
• LLM Critics Help Catch LLM Bugs (2024)

Honors

TIME100 Most Influential People in AI (2023, 2024)

Interviews and talks

• 80,000 Hours podcast #159 on superalignment (2023)80000hours.org