Artificial Intelligence for Humanitarian Response: Applications, Challenges, and Opportunities in Conflict and Crisis-Affected Regions A Review with a Case Study from Yemen

By the end of 2025, an estimated 117.8 million people worldwide were forcibly displaced by conflict, violence, and persecution, and humanitarian organizations face a persistent gap between the scale of need and the resources available to meet it. Artificial intelligence is increasingly deployed to narrow that gap: predictive models forecast famine and disease outbreaks before they escalate, satellite imagery paired with machine learning speeds up damage assessment, and AI-assisted translation connects aid workers with communities speaking under-resourced languages. This review surveys these applications and argues that crisis-affected regions pose design challenges distinct from typical AI deployment contexts unstable infrastructure, fragmented or contested data, and heightened privacy risk for already-vulnerable populations. These challenges are examined through the case of Yemen, where humanitarian data collection is complicated by fragmented territorial administration, yet where AI-assisted tools have supported disease-outbreak prediction and accountability in aid delivery. The review closes with recommendations for humanitarian AI that is co-designed with local and field expertise rather than imported unmodified from stable, high-resource settings.