Within the European Union Training Initiative (EUTI), a new research paper exploring the opportunities, structural risks, and regulatory implications of embedding Artificial Intelligence (AI) into EU Civilian Common Security and Defence Policy (CSDP) missions has been published.
Drafted by Dr. Francesco Paolo Levantino and Dr. Marta Stroppa from the Sant’Anna School of Advanced Studies (SSSA) within the framework of the EUTI Working Group on Innovation, the paper aims to contribute to current discussions on the responsible, effective, and human rights-compliant use of AI in civilian crisis management, in line with relevant EU and international legal and policy frameworks.
Shifting from Reactive to Anticipatory Crisis Management
As geopolitical instability, hybrid threats, and human rights violations intensify globally, civilian crisis management faces operational pressure. The paper outlines how AI systems can support a transition from reactive responses to a more proactive and adaptive framework.
Some of the insights from the Paper:
- The authors examine how Generative AI (GenAI) can move training beyond generic case studies. By utilizing frameworks like the BOPPPS model, instructors can leverage AI to instantly build role-specific scenario “injections”.
- Across the live mission lifecycle, AI can significantly improve situational awareness through satellite imagery and computer vision, reduce cognitive overload for human decision-makers, and optimize smart logistics or personnel matching.
- A core focus of the analysis is dedicated to technical limitations. The paper notes that fragile crisis environments often suffer from incomplete or biased datasets, warning that uncritical reliance on AI can trigger automation bias, deskilling, or create contextually plausible but factually flawed hallucinations.
- The integration of technological innovation must not occur in a normative vacuum. The paper highlights that under the EU AI Act, systems used to evaluate learning outcomes or guide vocational pathways in training environments are classified as High-Risk, legally mandating strict human oversight, logging, and data governance.
Actionable Recommendations
The publication concludes with seven foundational benchmarks aimed at guiding the European External Action Service (EEAS), the Civilian Planning and Conduct Capability (CivOpsHQ), and member states as they design official official EU operational strategies:
- Embed AI initiatives logically within existing CSDP training and operational policies.
- Utilize the EU AI Act, GDPR, and EUDPR as strict baseline quality benchmarks even where their direct application isn’t strictly mandatory.
- Conduct thorough, use-case-specific impact assessments (such as Fundamental Rights Impact Assessments) prior to any deployment.
- Prioritize comprehensive AI literacy across all personnel lines to preserve and reinforce meaningful human judgment over automated systems.
Read the full paper to explore the complete findings and policy recommendations: