Attributing Individual Criminal Responsibility for AI-Enabled Targeting of War Crimes: An Interpretive Framework Under the Rome Statute

Abstract

Artificial intelligence increasingly shapes military targeting through intelligence fusion, target classification, collateral-damage estimation and other decision-support functions. This article examines how individual criminal responsibility may be attributed when algorithmic outputs materially influence an allegedly unlawful attack. Focusing on Articles 25, 28, 30 and 32 of the Rome Statute, it argues that AI-enabled targeting does not create a general accountability gap, but complicates the identification of human conduct, mens rea , mistake of fact and command responsibility. It develops the Algorithmic Criminal Attribution Framework (ACAT), built around four operational rules: Actor–Output Nexus, Warning Specificity, Distributed Control and Temporal Reconstruction. These rules assess whether technical evidence—such as warnings, logs, confidence indicators and testing records—supports or undermines a legal element against a particular accused at the relevant time. ACAT creates no new basis of liability; it organises existing law while preserving individual culpability and proof beyond reasonable doubt.