When No Human Selects the Target: Individual Criminal Responsibility and Mens Rea for Fully Autonomous Weapons Under Article 25 of the Rome Statute

Abstract Autonomous weapon systems challenge international criminal law not because they create a new artificial offender, but because they fragment and temporally relocate human agency. This article examines individual criminal responsibility under Article 25 of the Rome Statute where a weapon selects and attacks targets without contemporaneous human approval. It treats the ‘culpability bottleneck’ not as a complete substitute for the broader accountability-gap thesis, but as the acute problem of connecting distributed human conduct and subjective culpability to a specific mode of liability. It argues that Article 25 remains applicable to many deliberate criminal uses, while recognising that some technological and organisational architectures may structurally disperse objective and subjective elements. The article develops a decision-chain model, clarifies the Article 30 boundary between knowledge and risk, and distinguishes predictable opacity, bounded stochasticity and radical malfunction. It also explains the complementary but limited roles of Article 28 and meaningful human control.