PhD, Postdoc, and Master's Thesis Opportunities in Technical AI Safety and Alignment ELLIS Institute Tübingen in Germany

Opportunity: Expression of interest for a new AI Safety and Alignment Group led by Maksym Andriushchenko at the ELLIS Institute Tübingen and Max Planck Institute for Intelligent Systems.

Open levels: The group is hiring PhD students, postdocs, master's thesis students, research interns, and research assistants. The postdoc call is for about 2 years and the start date is as soon as possible. PhD applicants can start in Fall 2025 or through centralized programs such as CLS, IMPRS, and ELLIS for Spring–Fall 2026.

Research areas: technical AI safety, AI alignment, frontier evaluations, scalable oversight, recursive self-improvement, autonomous LLM agents, AI risk measurement, capability evaluation, and generalization. The group emphasizes methods that are general and scale with intelligence and compute, and it is interested in both risks and capabilities.

Research style: The group values impactful work beyond traditional papers, including open-source repositories, benchmarks, blog posts, and other useful outputs. It explicitly mentions work on evals and methodological contributions, and highlights prior examples such as AgentHarm, OS-Harm, HalluHard, Skill-Inject, PostTrainBench, InferenceBench, random search attacks, prefilling attack, and Claudini.

Eligibility / application: The form asks for your current position and affiliation, CV, transcripts of records for bachelor's/master's students and PhD applicants, earliest start date, and short answers about research interests, favorite papers, achievements, and views on AGI and AI risks. No deadline is stated.

How to apply: Submit the Google Form expression of interest. If you are interested, you may also speak with Maksym Andriushchenko at ICML in Seoul. The form is intended to understand your interests and perspective, so concise answers are acceptable.