Mitt Regan (Georgetown University Law Center) has posted It Takes a System: AI-Enabled Weapons and the Law of War (Transnational Law & Contemporary Problems) on SSRN. Here is the abstract:
The international debate over the use of AI-enabled weapons has focused primarily on the presence or absence of human involvement in the selection and engagement of targets. This emphasis on the final moments of targeting is too narrow, because it neglects the fact that an AI-enabled weapon is a sociotechnical system: a set of humans, machines, and institutional rules and practices whose interactions produce the system’s outcomes.
This Article traces the history of the concept of a sociotechnical system and its application in various domains. It describes how the selection and engagement of targets is part of a sociotechnical system comprised of the targeting cycle, in which humans and AI-enabled decision-support tools interact across multiple stages. It then explains how the targeting cycle is nested with a larger sociotechnical system, and it is nested within the larger sociotechnical system of the AI weapon lifecycle.
Drawing on the framework first established by the International Electrical and Electronic Engineers (IEEE), the Article traces this lifecycle from deliberation before development through research and development, procurement, testing and evaluation, training, legal review, command and control, tactical employment, and post-deployment review. It also examines the ongoing activities that run across all of these stages. Each stage presents distinct opportunities, and distinct obligations, for humans to embed legal and ethical constraints in how a weapon operates. The 1988 downing of Iran Air Flight 655 by the USS Vincennes illustrates what this perspective adds. The tragedy resulted neither from simple “human error” nor from technological malfunction. It resulted from a failure in the interaction between operators and an interface whose design shortcomings had been identified, but not corrected, earlier in the Aegis system’s lifecycle.
The Article concludes that whether an AI-enabled weapon complies with international humanitarian law and broader moral norms depends on how the sociotechnical system functions as a whole, not simply on whether a human or a machine performs the final act of targeting. This requires greater attention to how humans and machines combine to perform functions across the AI weapon lifecycle, and their implications for responsible use of such weapons.
Recommended!
To receive new posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.
Lawrence Solum