AIC4ATM · Artificial Intelligence Compliance for Air Traffic Management solutions

The rapid adoption of Artificial Intelligence (AI) in Air Traffic Management (ATM) and airport operations offers efficiency gains but raises critical challenges of trust, safety, and regulatory compliance. The forthcoming EU AI Act and related frameworks (e.g., EASA AI) require AI systems in critical infrastructures to be explainable, auditable, and continuously monitored, standards lacking dedicated aviation-specific mechanisms.AIC4ATM addresses this gap by developing a modular AI Compliance Engine that integrates ethical safeguards, operational transparency, Explainable AI, and regulatory adherence for AI applications under development or approaching deployment. Combining top-down analysis of European legal frameworks with bottom-up insights from ATM stakeholders, including AI deployers such as airports, the project follows Responsible Innovation principles of anticipation, inclusion, ethics-by-design, and accountability.To achieve effective compliance, representative AI use cases in ATM and airport operations are identified and clustered, and their technical components decomposed to pinpoint obligations, building on the Common Taxonomy. These profiles populate a graph-based repository capturing interrelations among AI attributes, risks, and regulatory requirements, while processing the large data volumes typical of SESAR solutions. A risk taxonomy and compliance logic trees link identified risks to concrete obligations and support human–machine teaming.The AI Compliance Engine design provides end-to-end support, from risk categorization and automated compliance checklists to interfaces ensuring operational clarity and human-centric safeguards such as bias monitoring and continuous oversight. A development roadmap guides the concept to TRL 7 for real-world deployment. By enabling safe, certifiable AI use, AIC4ATM fosters dependable, data-driven innovation and drives digital transformation into one of Europe’s most safety-critical domains.