Abstract

As artificial intelligence (AI) systems are increasingly integrated into military operations, critical questions arise about the permissibility and feasibility of programming normative reasoning into AI systems. In this chapter, I address these issues in two main sections. First, I explore arguments that challenge the permissibility of delegating normative military reasoning to AI systems. I distinguish between concerns that stem from the characteristics of current AI technologies and the state of existing regulatory frameworks, and those that arise from fundamental moral and legal principles. The latter include inalienable obligations and activities that necessitate continuous human involvement. Second, I analyse the challenges of programming normative reasoning into AI, highlighting the constraints of rule-based and data-driven approaches. These constraints include the difficulties of modelling complex moral and legal reasoning, the risks associated with unreliable machine learning outputs, and the practical and ethical challenges of testing the reliability of high-stakes AI systems in real-world applications. In the face of these challenges, delegating normative military reasoning to AI systems remains impermissible and unfeasible.