Daniel Filan

Berkeley, California, United States

Bio

Updated 05/18/26

Daniel Filan is a member of technical staff at METR (Model Evaluation & Threat Research), where he works on assessing loss-of-control risk from frontier AI development. He completed his PhD in AI/EECS at UC Berkeley in 2024, supervised by Stuart Russell, with a thesis titled "Structure and Representation in Neural Networks"; prior to that he completed an honours year at the Australian National University supervised by Marcus Hutter, studying reinforcement learning theory, mathematics, and theoretical physics. From mid-2024 to late 2025 he was a senior research manager at MATS, helping train the next generation of AI alignment and interpretability researchers. He is the creator and host of AXRP (the AI X-risk Research Podcast, pronounced "axe-urp"), launched in December 2020, which features long-form interviews with researchers about their work on reducing catastrophic and existential risks from AI. He also hosts a sister podcast, The Filan Cabinet, covering a broader range of topics.