AI-driven war gaming as a tool for urban resilience: simulating conflict scenarios to enhance future city planning

Urban warfare remains a key blind spot in conventional planning paradigms, with cities exposed to cascading systemic collapse. In this research, we present AI-powered war gaming as an innovative method for simulating adversarial scenarios, ranging from attacks on infrastructure to mass displacement, and for strengthening urban resilience. Taking Mosul, Iraq, as a case study, we create a digital twin of the city and use agent-based modeling and adversarial AI to stress-test its post-war reconstruction. Experiments demonstrate that informal social networks (e.g. markets) outperform formal systems at crisis response, and that modular infrastructure design can reduce single-point vulnerabilities by 35–40%. However, the approach also exposes ethical hazards, including the possible weaponization of AI models and biases in training data. We make strategic policy recommendations to the United Nations Economic Commission for Europe (UNECE) and urban planners, highlighting the importance of fragmented redundancy in infrastructure networks, identifying organic nodes of resilience, and implementing strict humanitarian constraints on the use of artificial intelligence in cities. The implications conflict with traditional planning models that view conflict as an external shock, recommending its integration as a quantifiable element in resilience planning instead.