Chapter 8 looks at an Article 36 review of autonomous weapons systems. Having first addressed the progress of debates about potential future regulation of such systems at the international level, it delineates the notions of autonomy, artificial intelligence and learning systems. The chapter shows that military capabilities underpinned by machine learning and deep learning technologies require a new understanding of how provisions of targeting law can be meaningfully translated to the context of autonomous systems and examined in the framework of Article 36. It also demonstrates that determining the novel character of learning autonomous capabilities to ensure timely provision of legal advice, gaining assurance of complex adaptive systems’ performance accuracy and reliability, or identifying a set of requisite control measures, are all issues that ‘traditional’ weapons reviews do not have to grapple with at all, or at least not to the same extent. [143 words]

This chapter discusses the law governing the nature and design of weapons as it relates to autonomous weapon systems. It is presented in three parts. The first part covers prohibitions and restrictions on specific types of weapons. At present there is no law directed specifically at autonomous weapon systems, but it is conceivable that autonomous capabilities may be added to types of weapons that are subject to special regulation; the permissibility and effects of such development are considered. The second part discusses requirements imposed on weapons by general rules of IHL based on their nature and intended use. The fundamental principles of distinction and the avoidance of superfluous injury and unnecessary suffering are most at issue here. The main finding of this part is that an expanded understanding of ‘accuracy’ is required in order to effectively apply the principle of distinction to autonomous weapons. The third part discusses the Martens Clause, finding that, whether a broad or narrow interpretation of the Clause is adopted, it does not, at present, inhibit development or use of autonomous weapons.

This chapter links the distinguishing technological features of autonomous machines, identified in the previous chapter, to matters which are the subject of legal regulation: the decisions and actions of combatants, and the behaviour of weapon systems. It outlines the effects that weapon autonomy I likely to have on military decision-making processes and discusses the novel sources of risk and modes of failure associated with operating complex autonomous machines in combat. The problem of how to categorise autonomous weapons for legal purposes is addressed, as a response to suggestions by some commentators that weapon systems which are able to ‘step into the shoes’ of soldiers in some respects should perhaps be considered to be something more than just weapons when assessing their legal compliance. Finally, the conditions under which use of autonomous weapons may transgress the law are identified, as preparation for the detailed legal analysis to follow in the remainder of the book.