The first warm rays of the spring sun stream through the windows of the School of Computing, Information and Technology building at the Technical University of Munich (TUM). Groups of young researchers stand together, and a palpable excitement fills the air. For the next 24 hours, the familiar campus building at TUM will be their workplace. They have gathered to compete against one another in the fourth regional preliminary round of the CISPA European Cybersecurity & AI Hackathon Championship. Within a single day, they must solve three research-based challenges at the intersection of artificial intelligence and cybersecurity, with no possibility of an extension
Research Process Condensed into 24 Hours
After a welcome greeting from the host, Prof. Dr. Bernd Finkbeiner (TUM/CISPA), the competition began with a brief introduction by the championship’s initiators, CISPA-Faculty Dr. Franziska Boenisch and Dr. Adam Dziedzic. They present the three challenges the teams will be working on in the coming hours. Each is derived from real-world research problems currently being discussed in the AI and cybersecurity communities. Solving them requires both theoretical understanding and hands-on experimentation—much like the daily work of researchers. The tasks address topics such as model robustness and trustworthy AI, which are becoming increasingly important as AI systems become more prevalent in everyday applications.
From Initial Ideas to a Long Night of Coding
After the tasks are presented, the teams gather and retreat to individual workstations. Participants first get settled in before the actual work phase begins. Some participants start by reading academic publications on the topics of the tasks, while others test initial prototype models. Whether through a strategically planned approach or trial and error, there are many paths to the goal. As evening approaches, the pace shifts. A warm dinner provides a welcome boost. Conversations briefly move away from code and models and instead focus on studies, career paths, and experiences in computer science research.
A Night Shift as an Intensive Research Process
As darkness falls over Munich, the work continues. The leaderboard tracking the teams’ submissions is constantly changing as new solutions are uploaded and evaluated. Some teams take turns working in shifts to advance their projects and catch a few hours of sleep in between. Others choose to work through the night, repeatedly tweaking their code and retraining models. Parameters are fine-tuned, datasets reanalyzed, and new approaches tested. This intensity is what makes hackathons so special. In just 24 hours, participants experience a condensed version of the research process. They formulate hypotheses, test ideas, encounter obstacles, and try again.
Final push to crown the winning teams
Early in the morning, the fatigue in the room is clearly visible. But the decisive phase of the competition is still ahead: the final presentations. Here, the teams demonstrate how they analyzed the tasks, which methods they used, and the challenges they had to overcome. All team members must take the stage together, and no one is exempt from the jury’s critical questions. It’s not enough to have submitted a convincing solution; teams must also be able to defend it with solid reasoning. In the end, the teams Pentagon, Team ZOS, and CATByte impress the jury. Their approaches really stand out, qualifying them for the championship finals at CISPA in St. Ingbert in July 2026.
Two Keynotes as a Special Highlight
Unlike previous hackathons, the Munich event featured two keynote presentations. On Saturday, Prof. Dr. Stephan Günnemann from TUM spoke on “Reliable AI: Making Models Robust Enough to Survive Hackathons.” His talk connected fundamental research questions with the practical challenges participants face firsthand. On Sunday, Prof. Dr. Nicolas Papernot (University of Toronto) presented “LLM Dataset Inference: Did you train on my dataset?” He highlighted that the hackathon tasks build on foundations shaped by his own research. “Participants of the hackathon were very engaged with the topic of dataset inference and initiated many interesting discussions through their insightful questions. “Kudos to Franziska and Adam for organizing the event!”
Inspiration for Future Researchers
The keynotes once again underscored the championship’s goal beyond the competition itself: connecting young talent with the research community. Many participants are currently studying computer science, data science, or related fields. For them, the hackathon offered not only a hands-on insight into how research in AI and cybersecurity is applied in practice but also an opportunity for direct exchange with leading international researchers. This was made possible through the collaboration between CISPA and TUM, which hosted the fourth regional qualifier of the championship in Munich. The close connection between the competition and current research was further reflected in the nearly simultaneous IEEE Conference on Secure and Trustworthy Machine Learning (SaTML 2026) in Munich, co-chaired by CISPA-Faculty Lea Schönherr.
Our team in Munich
About the Hackathon Championship
The CISPA European Cybersecurity & AI Hackathon Championship is a Europe-wide competition organized by the CISPA Helmholtz Center for Information Security from November 2025 to July 2026. In regional qualifiers held in major European university cities, bachelor’s and master’s students compete in teams of up to four people to solve challenges in the fields of AI and cybersecurity within 24 hours. The winners from each city qualify for the grand finale in St. Ingbert, where they compete for cash prizes, trophies, and certificates. By bringing together young talent from across Europe, this championship aims not only to promote innovation and expertise in trustworthy AI and cybersecurity, but also to build a pan-European community dedicated to securing our digital future.
