CORPORATE CRIMINAL LIABILITY FOR MILITARY ARTIFICIAL INTELLIGENCE: PROSPECTS AND LIMITS OF AMENDING THE ROME STATUTE

Abstract

This study examines whether Article 25(1) of the 1998 Rome Statute, which confines the jurisdiction of the International Criminal Court to natural persons, can accommodate the accountability gap created by private corporations that design, test, and commercialise military artificial intelligence. It aims to determine whether extending jurisdiction to juridical persons is doctrinally coherent and institutionally feasible, and to specify the attribution standard such an extension would require. The study employs a normative-juridical method combining doctrinal analysis of the Rome Statute, read under Articles 31–32 of the Vienna Convention on the Law of Treaties, with a comparative survey of five domestic corporate-liability regimes and two illustrative episodes of corporate involvement in mass atrocity. The findings show that the existing modes of individual liability under Article 25(3) and Article 28, together with the doctrine of indirect perpetration through an organised apparatus of power, do not transfer cleanly to a commercial defence-technology enterprise, but that the principal obstacle to reform is not drafting inertia, as is commonly assumed, but the interaction of complementarity under Article 17, the principle of legality under Article 22, and the two-tier consequences of amendment under Article 121(5). The study concludes that a prospective Article 25 bis is legally defensible only if it adopts a non-derivative, failure-to-prevent standard of organisational fault, modelled on comparable domestic regimes, and is accompanied by a dedicated complementarity carve-out. This furnishes an attribution-specific and amendment-mechanism-specific model for corporate liability before the Court, moving the debate beyond a general case for reform toward an operable legal design.