At our Hackathon Headquarters at TecnoCampus in Mataró, Spain, temperatures are relatively cool, while outside the sun beats down hot on the asphalt. Inside the red brick building, with the sea just about 100 meters away as the crow flies, the teams are seated and, shortly after the competition begins, have their eyes and minds fixed solely on the tasks on their laptops. Anyone not involved in computer science or machine learning research sees only endless rows of numbers on the screens.
Cracking the Black Box
What are the teams struggling with? Challenges such as inferring sensitive characteristics of training data solely from a model’s predictions. “The participants are working with a black box. They have no idea how the model in front of them is structured. They’re developing a kind of shadow model to figure out, based on the answers, which models were trained on an unbalanced patient population. This means, for example, that the training dataset contains more data from men than from women. The teams never see the training data itself,” explains Aditya Kumar. He is a researcher at the SprintML Lab, which is led by the two scientific organizers, Franziska Boenisch and Adam Dziedzic, at CISPA. At the hackathon, Aditya is part of the support team that assists the participants.
Attacking to Understand
The purpose behind the competition’s challenges: Only when researchers understand where and how AI models and security mechanisms are vulnerable can they work specifically on methods that make AI models safer and more trustworthy. This is precisely the daily goal of Franziska, Adam, and their SprintML Lab. In his keynote at the start of the event, Jesse Cresswell of Layer 6 AI once again defined for the participants the characteristics AI must possess to earn our trust: According to the researcher, AI must be robust, fair, and explainable, and it must protect privacy in order to be trustworthy. Furthermore, he emphasized that it is important for models to be able to assess their own uncertainty and the accuracy of their decisions. He urged the participants to always view systems as a whole in their research, rather than focusing on individual aspects in isolation.
24 Hours of Full Concentration
But for now, the hackathon demands focus. Time marches on relentlessly, and the participants don’t even take the time to eat in peace. Tapas and pizza lie next to laptops that have gotten quite warm by now, forming strange works of art on the tables alongside coffee cups, banana peels, and energy drink cans. Adam explains: “Hackathons like this are so much more interesting than lectures or seminars because participants can immediately put the theory they’ve heard into practice.” Info sessions before the hackathon and close supervision during the event also help participants tackle the tasks. “You can tell that a team has really understood the task when it can clearly explain its approach and how it works. We go to great lengths to clearly define the goals, from the introductory session right through to the task descriptions. In the end, what makes the difference is how thoroughly the teams have thought through the problems.”
Turning Night into Day
Equipped in some cases with sleeping bags and a change of clothes, the hackathletes get ready for the night, during which most will end up sleeping little to nothing. At most a short “siesta,” as one participant explains to me with a tired wink. So it’s understandable that the participants’ eyes are small during the presentation of their results the next day and their faces look rather pale despite their spring tans. After everyone has presented their solutions, the jury briefly retires once more before the award ceremony signals the end of a long day for the competitors. First place goes to Miguel Requena Micó, Javier Polo Gambin, Domingo Méndez García, and Juan Luis Gomez Gomez from the team “CDL Chorrico de Limon”!
“It was chaotic. But in a good way.”
What do they take away from this hackathon besides the golden ticket to the finale? “In my studies, I tend to focus more on cybersecurity. I’ve learned a lot of new things here in the field of machine learning,” says Miguel. His teammate Javier agrees: “I feel the same way. It’s just that I usually focus on machine learning, and I found the tasks at the intersection of both fields very exciting.” Domingo admits: “For a long time, we really didn’t see ourselves as potential contenders for the win. At the beginning, we were still pretty low in the rankings. Then we worked our way up step by step.” It wasn’t until the final hour of the hackathon that they made decisive progress. What was the secret to their success? “We worked really well together because we know each other well by now. We study together at the University of Murcia in southeastern Spain and can also assess each other’s strengths and weaknesses. That allowed us to divide up the tasks very effectively,” explains Juan Luis. And Javier adds with a laugh: “It was chaotic. But in a good way.”
Our Team in Barcelona
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.
