Abstract
This paper asks whether Article 36 of Additional Protocol I functions as a genuine accountability safeguard for weapons review, or merely produces its appearance. Using doctrinal reconstruction of Article 36's architecture (via the ICRC's Guide and Commentary) combined with an interdisciplinary framework borrowed from Dan Davies' concept of the "accountability sink," the paper maps three structural features; delegation to process, broken feedback loops, and information-attenuating certification, onto State practice, supported by evidence of near-universal non-disclosure among States Parties. Autonomous weapon systems (AWS), illustrated through the 2020 Kargu-2 incident and analysed against the ICRC's "human control" standard and the Bemba command-responsibility jurisprudence, serve as the paper's aggravating case rather than its cause. The central argument departs from the dominant "accountability gap" literature: rather than diagnosing an absence awaiting correction through harmonisation or added process, the paper contends Article 36 is a functioning mechanism that converts the demand for accountability into its own discharge, such that AWS have not broken the mechanism but merely exposed its pre-existing design.