Maksym Andriushchenko

Maksym Andriushchenko is a principal investigator at the ELLIS Institute Tübingen and the Max Planck Institute for Intelligent Systems, where he leads the AI Safety and Alignment group. Recently, he served as chapter lead for the International AI Safety Report 2026 chaired by Prof. Yoshua Bengio. He collaborates closely with industry: he has participated in red-teaming efforts for OpenAI and Anthropic models, and the benchmarks he co-authored have been used by OpenAI, DeepMind, Meta, xAI, and Anthropic / UK AI Safety Institute. He obtained his PhD in machine learning from EPFL in 2024, supported by the Google and Open Phil AI PhD Fellowships. His PhD thesis received the ELLIS PhD Award and Patrick Denantes Memorial Prize at EPFL.

AI2050 Project

AI systems are becoming capable of performing AI research themselves: writing code, running experiments, and improving models with minimal human oversight. Andriushchenko’s project develops scientific tools to measure these capabilities and the risks they introduce. Using benchmarks like PostTrainBench, they will systematically evaluate how well AI agents can train other AI systems and whether they follow safety rules during this process. They have found that even today’s best AI agents sometimes cheat by using forbidden data or shortcuts. As AI approaches the ability to improve itself recursively, reliable measurement tools will be essential for keeping this technology safe and beneficial.