Document Type


Publication Date



National Security Law Program


Weapon systems are becoming increasingly automated and arguably some autonomous military systems have been deployed for years. Recent advances in automated systems and the possibilities they portend have generated interest and anxiety within some militaries and defense ministries, and a movement of non-governmental activists seeking to ban fully autonomous weapons. In May 2014, the High Contracting Parties of the UN Convention on Certain Conventional Weapons (CCW) convened an extensive discussion of the legal and ethical issues that autonomous weapons raise, while recognizing that many of these problems lie at an uncertain point in the future.

It is important that normative development regarding autonomous weapon systems head down a path that is coherent and practical. By "autonomous weapon systems," we mean systems "that, once activated, can select and engage targets without further intervention by a human operator." We draw this definition from a 2012 U.S. Department of Defense policy directive, which remains the most extensive public pronouncement by any State on how it intends to proceed with regard to research, development and deployment of autonomous weapon systems.

This paper addresses several questions that are critical to charting such a path. First, are autonomous weapon systems different from other new weapon systems, and, if so, how? Second, to the extent they are different, can and should autonomous weapon systems be regulated within the framework of the existing law of armed conflict? If yes, how should States go about doing so? If not, what alternative regulatory approach is appropriate?

We conclude that autonomous weapon systems have special features that pose risks and that create challenges in applying the existing law of armed conflict. Nevertheless, we conclude it is possible to adapt the existing framework to account for the features of autonomous weapons, and that the suggested alternative of prohibiting these systems outright is misguided. Instead, we propose a three-tiered process for regulating the development, deployment and use of autonomous systems.