Artificial intelligence (AI) is transforming how natural and anthropogenic hazards are monitored, modelled and governed, from early warning to scenario-based planning. This session invites contributions that critically examine both opportunities and limitations of AI across the hazard sciences. Example topics include representative and open datasets (including small, sparse or biased data), uncertainty quantification and communication, explainability and accountability, multi-hazard and cascading applications, and operational deployment with practitioners. We particularly welcome work that: (i) demonstrates reproducible methods and rigorous evaluation; (ii) addresses sustainability concerns (compute, energy and water footprints) and ethical equity; and (iii) translates models into decisions under real-world constraints at different decision-making levels. Policy- and practice-oriented perspectives are also encouraged, including analyses of how EU initiatives and funding programmes (e.g. the Union Civil Protection Mechanism, “Preparedness Union”) and the evolving EU AI regulatory framework can support trustworthy, fit-for-purpose AI in risk management. Case studies, benchmarks, negative results and dataset papers are likewise welcome.