ABSTRACT
Risks posed by different military systems which leverage Artificial Intelligence (AI) technologies may vary widely and applying common risk-mitigation measures across all systems will likely be suboptimal. Therefore, a risk-based approach holds great promise. This chapter presents a qualitative model for such an approach, termed the Risk Hierarchy, which could be adopted for evaluating and mitigating risks posed by AI-powered military systems. The model evaluates risks based on parameters which adequately reflect the key apprehensions emerging from AI empowerment of military applications, namely, violation of IHL and unreliable performance on the battlefield. These parameters form the basis for mapping the wide spectrum of military applications to different risk levels. Finally, in order to mitigate the risks, modalities are outlined for evolving a differentiated risk-mitigation mechanism. Factoring in military ethos and analysing risks against the backdrop of realistic conflict scenarios can meaningfully influence risk evaluation and mitigation mechanisms. The rigor which underpins the Risk Hierarchy would facilitate international consensus by providing a basis for focussed discussions. The chapter suggests that mitigating risks in AI-enabled military systems need not always be a zero-sum game, and there are compelling reasons for states and militaries to adopt self-regulatory measures.
