As the use of AI expands across civil protection, it also brings a set of practical and ethical challenges that must be addressed for these tools to function reliably in real emergencies. Questions around trust, explainability, data integration, and procurement are now central to discussions taking place across Europe. The following section outlines some of the main issues shaping this debate and what they mean for the responsible adoption of AI in disaster risk management.

Limits and lessons: challenges in bringing AI into practice

Despite clear progress, bringing artificial intelligence into real-world disaster management remains a complex task. AI systems can extend human capability, but their adoption depends on trust, transparency, reliable data, and close alignment with operational needs. The following issues are central to ongoing discussions across Europe.

Trust, explainability & human oversight: In life-saving situations, decisions based on opaque algorithms raise serious concerns. Civil protection authorities must be able to understand and explain the recommendations made by AI systems, especially when they guide real-time operations. “Socially and ethically, there are concerns about trust, transparency, and the acceptance of robots in roles traditionally filled by humans, particularly in sensitive situations involving victims.”— Alexandre Ahmad, CARMA

Training institutions such as ENSOSP stress that explainable AI (XAI) is as much about human confidence as technical design. If AI outputs are difficult to interpret, they risk adding to rather than reducing responders’ workload. “AI can support human decision-making, but it cannot replace professional judgement.” — Quentin Brot, ENSOSP

CARMA’s experience reinforces this lesson: human–robot collaboration depends on transparency and predictability. When users can understand how a system reacts ,and why, they are far more likely to trust and adopt it in high-risk environments. “Trust will come through transparency — responders must understand not just what an AI tool tells them, but why.”— Dr Nicholas Vretos, CARMA

Data, integration & interoperability: A major challenge for AI in disaster management is how data from different sources — satellites, drones, sensors, and field reports — is gathered and shared. Effective use of AI depends on clean, harmonised, and interoperable datasets, but information is often fragmented or collected under pressure.

Projects such as OroraTech, ICEYE, and SYNERGISE highlight the difficulty of combining these inputs into one operational picture. “AI can enhance response capacity — but when communications fail or systems don’t connect, its advantages quickly diminish.” — Dr Sabina Ziemian, SYNERGISE

The need for standardised data protocols and shared interfaces is now widely recognised across Europe’s civil protection community. Without them, even the most advanced AI tools risk staying confined to pilot projects rather than becoming part of daily operations.

From prototype to procurement: Many AI tools in disaster management remain at the pilot stage — tested successfully but not yet part of routine operations. Procurement rules, certification processes, and short funding cycles often slow progress.

Developers and responders agree that innovation needs to move faster and stay close to user needs. “It takes time to build confidence in new technology. Users need proven accuracy, reliable monitoring, low false-positive rates, and transparent data before committing to large-scale use.” — Julia Gottfriedsen, OroraTech

Bringing AI from prototype to practice will require closer collaboration between researchers, policymakers, and practitioners, ensuring that new tools are practical, trusted, and ready for real-world use.