A Scoping Review of the Ethics of AI in Humanitarian Contexts: Principles, Governance, and the Path to Practice

Artificial intelligence (AI) is increasingly used in humanitarian action, yet raises ethical concerns about bias, transparency, privacy, and accountability. Ethical principles, governance mechanisms, and operational tools have each received attention, but how they relate, and what evidence exists on their implementation, remains unclear. We conducted a scoping review following the Arksey and O’Malley framework, as refined by Levac et al., reported it according to PRISMA-ScR, and registered it retrospectively on the Open Science Framework. Six bibliographic databases, Google Scholar, and targeted grey-literature sources were searched for documents published in 2015–2025, and two reviewers independently screened and extracted data. Thirty-two documents (15 peer-reviewed publications and 17 grey-literature documents) were included. Themes on fairness, non-discrimination, and inclusion (100%) and on transparency, explainability, and accountability (93.8%) were most common; grey-literature documents addressed more themes than peer-reviewed publications (mean 6.4 vs. 5.0 of 8). Most documents addressed ethics at a procedural level (71.9%), 21.9% addressed ethics at an operational/implemented level, and 31.3% reported implementation evidence. Only 12.5% mentioned the Middle East and North Africa. These findings describe what documents report, not what organizations do in practice.