Evžen Wybitul
Bio
Evžen Wybitul is a Czech AI safety researcher and Data Science Master's student at ETH Zürich. He completed his undergraduate studies in Bioinformatics at Charles University in Prague, graduating top of his class. He participated in the MATS (Model-Alignment Training for Safety) program in Berkeley, where he worked on scalable oversight for reinforcement learning and co-authored the paper "Gradient Routing: Masking Gradients to Localize Computation in Neural Networks" (arXiv 2410.04332). He has also collaborated with DeepMind's David Lindner on benchmarking safety-relevant capabilities of vision-language models. His research interests include mechanistic interpretability, scalable oversight, and AI governance; he is the sole author of "Access Controls Will Solve the Dual-Use Dilemma" (arXiv 2505.09341), accepted at the ICML 2025 Workshop on Technical AI Governance.
Links
- Personal Website
- https://evzen.dev/
- Twitter / X
- LessWrong
- -
Grants
No grants recorded.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Mar 22, 2026, 3:56 PM UTC
- Created
- Mar 20, 2026, 3:00 AM UTC