This report was written by Eva-Nour Repussard.
Image credits: U.S. Army Reserve photos by Spc. William Kuang, via DVIDS
Militaries are rapidly integrating artificial intelligence (AI) into operational and strategic decision-making.
AI can process information at a scale and speed beyond human ability, enhance situational awareness, support targeting and intelligence analysis, and assist decision-making in time-critical environments.
However, these capabilities also create significant operational, legal, ethical, and strategic risks, including cognitive bias, system opacity, adversarial manipulation, unreliable outputs, and inappropriate reliance on AI-generated recommendations.
Central to both these opportunities and risks is the relationship between humans and machines. The key governance challenge is not simply whether humans remain involved in military decision-making, but whether that involvement enables meaningful judgement and control. An operator may formally remain part of a decision-making process while lacking the information, time, authority, or training necessary to critically assess an AI-generated recommendation. Human judgement is itself imperfect: military personnel bring assumptions, cognitive biases, experiences, and perceptions of risk to their interaction with AI systems. Effective military AI governance must therefore address both the limitations of AI systems and the ways in which humans interpret, trust, challenge, and act upon their outputs.
Drawing on 32 semi-structured interviews with stakeholders across 14 countries, alongside policy and academic research, this report argues for a context-sensitive and risk-proportionate approach to Human-Machine Interaction (HMI). Rather than prescribing a uniform level of human involvement across all military AI applications, the appropriate form of oversight and decision authority should reflect the operational context, the level of uncertainty, the consequences of error, and the time available for human assessment. Current AI systems remain vulnerable to error, adversarial manipulation, and performance degradation when operating outside the conditions in which they were developed and tested. The report focuses on how AI can augment human capabilities and improve military decision-making while ensuring that human involvement remains meaningful. Decisions involving the use of lethal force should remain with an appropriately authorised human decision-maker.
Read the report and policy recommendations below.
