For the third time, the ELSA network gathered for the ELSA General Assembly. We met from June 17th – 19th in beautiful Brussels. Over three intense days, we gained insight into research progress in various fields, the importance and strategy of ensuring safe and secure AI for Europe, the challenges of disseminating research results to political stakeholders, and the evolving AI research landscape in Europe.
We enjoyed the posters and presentations from industry partners, researchers, as well as the winners of our second industry call. We learned about the latest progress and opinions on the future of research from representatives of the European AI Office, the European Commission, the Helmholtz Association, and other Networks of Excellence.
About the General Assembly
Every year, we invite the ELSA network, related parties, and stakeholders to the ELSA General Assembly. With participants representing politics, research, and various industries, the event offers a unique opportunity to discuss the progress of safe and secure AI research from multiple perspectives. Participants can gather unique insights, strengthen their private network, and find synergies.
This year, we were delighted to also welcome two of the winning SMEs from our first industry call. Representatives from QuantPi and OASYS NOW gave a talk and shared their valuable experiences with our network and the winners of our second industry call.





Presentations, posters, and proud award winners
Bridging the gaps between research, politics, and industry is a central point of the ELSA network. This is why we included a broad range of topics during the presentations.
Our network exchanged actions, ideas, and progress, i.e., about the European Commission’s plans on Artificial Intelligence (AI), Trustworthy and Safe AI, the connections between policy visions to scientific practice, the progress of the ELSA Use Cases and Industry Call winners, deface detection, real-world AI agents, robust Machine Learning, challenges towards trustworthy AI in industry, and the accessibility of personalized health.
This variety was made possible by the numerous speakers from high-ranking institutions, SMEs, industry, and NoE representatives. We want to thank Detesia, Invariant Labs, UniCa, Pluribus One, QuantPi, OASYS NOW, the European Commission, the EU AI Office, the Representation of Saarland to the European Union, CISPA, and Helmholtz Society Brussels, Delft University of Technology, ELIAS, ELLIOT, the University of Modena and Reggio Emilia, ENFIELD, dAIEDGE, ETH Zurich, and many more for their contributions.
Best Poster Awards
With an impressive 24 posters and 17 pitches, there was a wealth of new research to discuss and learn from. The topics ranged from certifiability, benchmarking, and differential privacy to Document Visual Question Answering (DocVQA), malware detection, machine learning, large language models, and many more.
The overall high level of research quality made it especially hard for the jury to select the three best poster award winners. On June 19th, Mario Fritz announced the jury’s decision:
- Alaa Anani (MPI/CISPA) for her poster and pitch about
“Pixel-level Certified Explanations via Randomized Smoothing” - Mackenzie Jorgensen (The Alan Turing Institute) for her poster and pitch on
“Documenting Deployment with Fabric: A Repository of Real-World AI Systems and Governance Mechanisms” - Xin Chen (ETH Zurich) for her poster and pitch on
“Learning Safety Constraints for Large Language Models” (ICML 2025 Spotlight)
Congratulations to our winners!



Industry Call 2: Winners showcase their progress
We emphasize the importance of connecting industry and research. Hence, the seven winners of our second Industry Call used the opportunity to present their progress:
- Kiin Bio talked about “Disrupting Drug Discovery with Virtual Scientists.”
- LightMirror presented their “Metamodel to streamline building–energy renovation planning.”
- ReqCon elaborated on “Increasing Explainability of ReqCon’s AI Tools for Construction.”
- Starlab Barcelona SL introduced “Fair decisions in neurodegenerative disease diagnostics.
- Trace Labs (OriginTrail core developers) shared their “Decentralized Genomic Data Sharing for Personalized, AI-driven Healthcare.”
- VERTLINER presented “RL-Based UAV Control for Infrastructure Inspection and Mapping”
- Vivid Mind talked about “Screening for the Early Onset of Dementia Using Voice.”
All companies showcased excellent developments and connections between our research network and their businesses.
On LinkedIn, Bluesky, and via our newsletter, we will continue to share the success and developments of our IC2 winners.



Andreas Withaeckx, ReqCon




Active participatioN during Workshops
Another highlight was the workshops. On day two, participants dove into the following topics:
- AI Robustness and Safety for Cybersecurity
- Towards Comprehensive Benchmarking for Privacy-Preserving Synthetic Biological Data
- Trade-offs in Trustworthy AI
- OriginTrail Decentralized Knowledge Graph (DKG)
We thank Battista Biggio, Maura Pintor, Fabio Brau (University of Cagliari), Jurij Skornik (Trace Labs), Mario Fritz, Ruta Binkyte (CISPA), and Hakime Öztürk (EMBL) for sharing their knowledge and leading the workshops.





Big news: Announcing the Extension of ELSA
On July 17th, ELSA Coordinator and CISPA-Faculty Prof. Dr. Mario Fritz announced the extension of ELSA. Our network will be able to continue its outstanding work and research for an additional year until the end of August 2026.
Read about the ELSA extension here: ELSA Extension Press Release.
