Automation is a key technical lever for enabling the Digital European Sky. Automating complex, safety-critical tasks requires artificial intelligence, particularly machine learning, yet traditional development methods do not address the specific challenges of AI-based systems. MEDUSA develops a model-based framework for the design of safe, secure and transparent-by-construction AI systems, aligned with data and model requirements defined in EASA concept papers.

The approach is based on operational domain models derived from use cases, which are used to generate representative training and test scenarios. Explainability and robustness techniques assess model behaviour and feed back into model refinement, improving safety and trust. Complemented by a model-based security approach, MEDUSA delivers structured assurance cases, demonstrated through representative air traffic management use case