Summary
Sahar Abdelnabi is inviting applications for 3–4 PhD positions, research visits, and postdoctoral opportunities within the COMPASS research group at the ELLIS Institute Tübingen and the Max Planck Institute for Intelligent Systems, Germany.
The research group, COMPASS (COoperative Machine intelligence for People-Aligned Safe Systems), focuses on AI safety, security, alignment, autonomous AI agents, and developing advanced AI systems that are aligned with human and societal values.
The group welcomes interdisciplinary research spanning computer science, social science, economics, and cognitive science, with strong connections to industry and governmental AI safety research institutes.
PhD, Research Visit & Postdoctoral Positions in AI Safety and Alignment at ELLIS Tübingen / Max Planck Institute
Summary Table
| Particular | Details |
|---|---|
| Positions | PhD Students, Research Interns (Predoctoral), Visiting PhD Students, Postdocs |
| Number of Openings | 3–4 PhD positions; additional research visits/postdoc opportunities |
| Research Group | COMPASS – Cooperative Machine Intelligence for People-Aligned Safe Systems |
| Research Areas | AI Safety, AI Alignment, AI Security, Autonomous Agents, Interpretability, Multi-Agent Systems |
| Institution | ELLIS Institute Tübingen / Max Planck Institute for Intelligent Systems |
| Location | Tübingen, Germany |
| Eligibility | Master’s degree for PhD applicants; relevant AI/ML research exposure and strong research motivation |
| German Language | Not required |
| Application | Submit the research-interest/application form; PhD applicants may also apply through ELLIS, CLS, or IMPRS-IS |
| Application Deadline | Not specified; applications are being considered for multiple openings |
Designation
- PhD Student
- Predoctoral Research Intern
- Visiting PhD Student
- Postdoctoral Researcher
Research Area
The positions are centered on AI Safety, Security, and Alignment, particularly the safety of advanced autonomous AI and future AGI systems.
Potential research directions include:
- Contextual AI agents that respect privacy and safety norms
- Red-teaming and securing real-world AI systems
- AI agents with long-term memory
- Self-evolving and open-ended agents
- Cooperative multi-agent systems
- Interpretability and white-box control methods
- Evaluation of emerging AI capabilities
- Deceptive alignment and situational awareness
- Long-term interaction and gradual loss of control
- AI for mediation and finding common ground
- AI for scientific discovery
Location
Tübingen, Germany
The research environment provides extensive opportunities for collaboration with researchers working in AI and AI safety. The group also maintains and plans to strengthen collaborations with industry research teams.
No knowledge of German is required.
Eligibility / Qualification
For PhD Applicants
Candidates should:
- Hold a Master’s degree or equivalent qualification.
- Be passionate about AI Safety and interested in learning about emerging AI safety problems.
- Be familiar with current AI research through activities such as:
- MSc thesis
- Research seminars
- AI Safety Camp
- Conferences
- Course projects
- Relevant research experience
- Demonstrate strong motivation for conducting impactful and foundational research.
A publication record is welcome but not mandatory. Passion, curiosity, research potential, and willingness to learn are particularly valued.
Other Applicants
The group is also considering:
- Predoctoral research interns
- Visiting PhD students
- Postdoctoral researchers
Specific requirements for these positions may vary depending on the research project and candidate profile.
Job Description
Selected researchers will contribute to the broader goal of ensuring that the next generation of advanced and potentially super-intelligent AI systems remains aligned with human and societal values.
Research will focus on proactively identifying and addressing the safety implications of increasingly capable autonomous AI systems while remaining grounded in practical challenges facing today’s AI systems.
The research environment emphasizes:
- Foundational and impactful AI safety research
- Understanding fundamental limitations of advanced AI
- Developing new perspectives and methodologies
- Rigorous evaluation of AI capabilities and risks
- Designing safer and more aligned autonomous agents
- Interdisciplinary collaboration
- Collaboration with academic, industrial, and governmental AI safety organizations
PhD Admission Channels
PhD applicants may also pursue admission through:
- ELLIS PhD Program
- Max Planck & ETH Center for Learning Systems (CLS)
- International Max Planck Research School for Intelligent Systems (IMPRS-IS)
Candidates with a Master’s degree who are interested in pursuing a PhD can apply directly through the advertised application form.
Research Environment & Highlights
- Excellent AI research environment in Tübingen
- Strong focus on AI safety and alignment
- Opportunities for collaboration with leading researchers
- Ongoing and planned collaborations with industry research teams
- Interdisciplinary research across multiple fields
- No German language requirement
- Opportunities to work on emerging and high-impact AI safety challenges
How to Apply
Interested candidates should complete the application/interest form provided by Sahar Abdelnabi to express their interest in working with the COMPASS research group.
PhD candidates can additionally explore the ELLIS PhD Program, CLS, and IMPRS-IS admission routes.
Applicants should highlight their academic background, relevant research experience, interests in AI safety/alignment, and motivation for pursuing impactful research.
Due to the large number of applications, responses may take time, and individual responses cannot be guaranteed. Candidates will be contacted if their profile is considered a good fit.
Last Date for Apply
Until position filled
Applications are currently being considered for the available positions, so interested candidates are encouraged to apply as early as possible.
About the Research Group
Sahar Abdelnabi is a Principal Investigator at the ELLIS Institute Tübingen and an independent research group leader at the Max Planck Institute for Intelligent Systems. She leads COMPASS, focusing on cooperative machine intelligence, AI safety, security, and alignment.
The group’s research includes influential work on indirect prompt injection attacks and generative AI watermarking, with research that has also received major recognition, including a Best Paper Award at ACL 2025.
