At the Crossroads of Control: The Intersection of Artificial Intelligence in Autonomous Weapon Systems with International Humanitarian Law

Abstract

This Article explores the interaction of artificial intelligence (AI) and machine learning with international humanitarian law (IHL) in autonomous weapon systems (AWS). Lawyers and scientists repeatedly express a need for practical and objective substantive guidance on the lawful development of autonomy in weapon systems. This Article proposes five foundational principles to enable development of responsible AWS policy. The findings emerged from a research project conducted by a team of military and civilian professors at the Stockton Center for the Study of International Law at the U.S. Naval War College. The study is informed by experts in computer sciences, government and military, non-governmental organizations, think tanks, and academia.

Advances in AI will likely produce AWS that are different in kind from existing weapon systems and thus require a fresh approach to evaluating IHL compliance. First, this Article describes the technological details pertinent to understanding the distinction between current and future systems. It argues that the technological evaluation of the spectrum of autonomy should focus on the combination of authorities granted to the computer that controls an AWS, while also taking into account the physical capabilities of the system. Second, it argues that a key issue bearing on IHL compliance is whether an AWS has been granted some combination of authorities and capabilities that functionally delegate the decision to kill from human to machine. Third, it posits that predictability must be at the core of an evaluation into whether a particular AWS breaches this delegation threshold and examines how AI handles uncertainty, a critical component of the predictability analysis. Finally, the Article proposes five foundational principles to guide the development of AWS policy.

Keywords: autonomous weapon systems, autonomy, international humanitarian law, artificial intelligence, machine learning