Obstacles to Criminal Liability: A Systemic Analysis

The criminal responsibility gap with respect to autonomous weapons (AWS) asserts that one or more elements required for criminal liability cannot be established due to the the technology’s unique properties. To explore the validity of this hypothesis, this chapter starts by providing an overview of the different elements required for ciminal liability. It finds that mens rea and actus reus are universally applicable, but that some theories also add control and causality as necessary elements. The chapter first examines the latter two, finding that the significant influence exerted by the commander during the deployment decision is sufficient to establish both causality and control. In contrast, this is less easy to determine with respect to persons earlier in the lifecycle (e.g. programmers), due to the Problem of Many Hands. The actus reus element is analysed next: it is argued here that for AWS, the human decision to activate the system under the chosen conditions should be considered as the actus reus, instead of the (autonomous) activities of the AWS in the field. Finally, with regard to mens rea, it is found that this element presents the most challenges: humans both at the operational or preparatory stages will often experience reduced perceivability due to the complexity of the AI models, their opacity, etc.—this epistemic problem makes establishing mens rea difficult if higher degrees of cognition are required for prosecution. The chapter concludes by identifying mens rea as the element that is most under pressure by the unique properties of AI.