AI systems reshape society in profound ways. To harness their potential, we need to prevent catastrophic risks through robust approaches to AI safety and governance that address questions of control, power, fairness, and societal impact.
We bring together curious and impact-driven minds across disciplines at Saarland University and beyond to tackle these challenges through research, education, and collaboration.
Learn how to join us:
Advanced Fellowship
Go deeper into the technical methods shaping the future of AI safety through a focused reading, discussion, and hands-on program:
Evaluation: learn how we test what capabilities AI models have and how they will act in certain situations
Interpretability: Explore what we can learn about AI models by looking at their internal workings
Engage with active research: discuss key papers, run hands-on experiments, and hear from researchers working on these problems
For those ready to move beyond the fundamentals and work on the problems that matter most.
Find out more here.
Introductory Fellowship
Learn the fundamentals of AI Safety and Governance through a weekly reading and discussion group:
Explore the key challenges: technical safety, governance, power dynamics, and societal impact
Discuss current approaches to ensuring AI systems remain controllable and beneficial
Build understanding in a supportive, interdisciplinary community
Your first step toward understanding and contributing to one of the defining challenges of our generation.
Find out more here.
AI Security Team
Making AI safe also requires security to avoid adversaries gaining access to powerful models, individuals training secret loyalties into models or models exfiltrating themselves.
We are a new sub-chapter of AISS looking deep into those questions. Join us if you are especially interested at the intersection of cybersecurity and AI!
-
We are hosting a 6-week closed, in-person reading group exploring critical and neglected problems in AI security. Based on recent research frameworks, this series will dive into the most pressing threat models and proposed research agendas for safeguarding advanced AI systems.
-
End of May we are aiming to host an afternoon filled with talks from experts in AI Security and plenty of chances to network and learn how you can contribute to securing powerful AI models.
-
We are currently in the process of planning a dedicated AI security research sprint, where you can make progress and gain experience on important security research questions. Stay tuned for further announcements!
Upcoming events.
We host regular public events with invited speakers as well as casual get-togethers to foster exchange and build a community.
As part of our flagship Interdisciplinary Research Incubator and Introductory Fellowship, we offer further exclusive events to our fellows.