AWS and Targeting

This chapter engages in a thorough examination of the five main targeting rules in IHL (indiscriminate attack, proportionality, minimising collateral damage, warnings, and delaying/cancelling) and how their application would be impacted by the unique properties of autonomous weapon systems (AWS). The chapter offers a dual approach: first, an analysis of how AWS-users should apply these obligations under a more strict interpretation of the law, and second, it considers a slightly more normative angle based on an object and purpose interpretation. This latter prong is motivated by certain characteristics of AI that can be capitalised on to improve the protection of civilians, given the willingness to read existing rules through a more teleological lens. Several new and important concepts are presented related to AWS-attacks. The notion of Attack Scopes is introduced to limit how broadly AWS-attacks may be construed under IHL: this subsequently is used to demonstrate how proportionality and cancelling would be performed for AWS-attacks. Related to indiscriminate attack, it is argued that deploying an AWS in circumstances outside of its intended use-cases or without (human-performed) target validation should be considered an indiscriminate attack, even if the system itself is not indiscriminate. In addition, it is argued that all Context Controls discussed in Chap. 8 constitute forms of risk mitigation, that are obligatory under the precautionary rule. This chapter thereby offers an extensive and reasoned view on how commanders should navigate targeting rules when using a novel technology for which minimal guidance currently exists.