Jan Leike

German AI alignment researcher; currently leads the Alignment Science team at Anthropic, working on scalable oversight, weak-to-strong generalization, and automated alignment research. He holds a PhD in machine learning from the Australian National University (supervised by Marcus Hutter) and completed a postdoctoral fellowship at the Future of Humanity Institute.

From 2016 to 2021 he worked in DeepMind's AI safety group, where he co-authored the 2017 paper "Deep Reinforcement Learning from Human Preferences" — early empirical groundwork for reinforcement learning from human feedback (RLHF). He joined OpenAI in 2021, leading the alignment team and contributing to InstructGPT (2022), the instruction-following model that brought RLHF into large-scale language models.

In June 2023 OpenAI named Leike and Ilya Sutskever co-heads of a newly formed Superalignment initiative, announcing a commitment of 20% of compute and a four-year timeline to solve alignment for superhuman AI systems. On 17 May 2024 he resigned — two days after Sutskever's own departure — publishing a thread on X stating that "safety culture and processes have taken a backseat to shiny products." OpenAI subsequently confirmed the Superalignment team's dissolution. The departure is documented in the insider whistleblowing from frontier labs strategy entry as one of the precipitating events behind OpenAI's pre-emptive reversal of its nondisparagement agreements and the Right to Warn open letter of 4 June 2024. He joined Anthropic later that month. He was named to TIME's list of the 100 most influential people in AI in both 2023 and 2024.