Who Acts When Autonomous Weapons Strike?

Abstract

This essay examines the theories according to which ‘actions’ carried out by autonomous weapon systems enabled by strong artificial intelligence in detecting, tracking and engaging with the target (‘intelligent AWS’) may be seen as an ‘act’ of the weapon system for the purpose of legal responsibility. The essay focuses on the material act required for the commission of war crimes related to prohibited attacks in warfare. After briefly presenting the various conceptions of the act as an essential component of the material element of criminal offences, it argues that the material act of war crimes related to prohibited attacks is invariably carried out by the user of an ‘intelligent AWS’. This also holds true in the case of so-called ‘unintended engagements’ during the course of a military attack carried out with an intelligent AWS. The essay moves on to examine the question of whether, in the case of the use of intelligent AWS by the armed forces of a state, the ‘actions’ of intelligent AWS — including those not intended by the user — are attributable to the state. It demonstrates that under a correct understanding of the concept of ‘act of state’ for the purpose of attributing state responsibility under international law, such attribution is unquestionable. It underlines that, suggesting otherwise, would bring to a breaking point the possibility of establishing violations by states of international humanitarian law in the conduct of hostilities.