State Reliance on AI for Humanitarian Activities in Armed Conflict
Framing Initial IHL-related Considerations
Dustin A. Lewis and Taha Wiheba • HLS PILAC • July 2026
Executive Summary
As artificial intelligence (AI) capabilities continue to develop, multilateral discussions concerning AI and armed conflict have focused primarily on military applications, including autonomous weapon systems, targeting, cyber operations, and battlefield decision support. Far less attention has been devoted to a related question: whether States may appropriately rely on AI in carrying out humanitarian activities regulated by international humanitarian law (IHL).
This analysis seeks to encourage government legal advisers and humanitarian-policy officials to begin developing views on that question before political, economic, or operational pressures might constrict opportunities for careful reflection. Although relatively few publicly documented examples currently involve extensive State reliance on AI for humanitarian activities, rapidly evolving capabilities may increasingly generate proposals to employ AI in areas such as needs assessments, humanitarian-relief planning, civilian protection, displacement forecasting, and the allocation of scarce resources.
This paper suggests evaluating such proposals not only by assessing whether AI may improve efficiency, speed, or consistency. States should also consider whether reliance on AI would affect the implementation of applicable IHL obligations as well as the achievement of intended humanitarian purposes. Humanitarian activities frequently require context-sensitive assessments concerning vulnerability, dignity, access, prioritization, and protection. These judgments often depend on local knowledge, professional expertise, and human responsibility that may not readily be captured through AI outputs alone.
To support such assessments, this paper applies HLS PILAC’s Exercising Cognitive Agency framework. The framework proceeds from the premise that IHL assigns responsibility for administering IHL obligations to legally responsible humans acting on behalf of a party to the conflict, not to machines. It therefore asks whether a legally responsible human remains capable of exercising sufficient cognitive agency when relying on AI to assist in executing cognitive tasks required by IHL.
The analysis proposes five initial considerations for government decision-makers: (i) identifying the relevant IHL obligations and humanitarian objectives; (ii) determining the cognitive tasks those obligations entail; (iii) assessing the role that AI would play in executing those tasks; (iv) evaluating whether reliance on AI would support, alter, or undermine the humanitarian character and objectives of the activity; and (v) determining whether a legally responsible human would be capable of exercising sufficient cognitive agency while relying on the particular AI tool or technique at issue. Illustrative examples concerning humanitarian-relief operations and AI-based identity verification demonstrate how these considerations may apply in practice.
The analysis does not seek categorical conclusions regarding potential AI use. Instead, it offers a structured framework to help States evaluate whether — and, if so, under what conditions — reliance on AI may or may not assist in upholding respect for IHL and in supporting the humanitarian purposes of relief and protection activities.