치명적 자율무기체계(LAWS)의 공법적 규율 방안 연구- 결심의 은폐와 가시화를 중심으로 -

The emergence of Lethal Autonomous Weapon Systems (LAWS), which autonomously select and engage targets without human intervention, has precipitated a fundamental crisis in the structure of responsibility attribution for state power exercise. While the exercise of state power to deprive life inherently requires human moral judgment and legal accountability, LAWS undermines this constitutional foundation by delegating lethal decisions to algorithms. Existing scholarship has proposed enhancing algorithmic explainability (XAI) or establishing Meaningful Human Control (MHC) standards, yet the former merely clarifies decision processes without identifying responsible agents, while the latter remains an abstract guideline lacking legal enforceability. This study adopts Steven Lukes' three-dimensional power theory andJohn Gaventa's power cube model as analytical frameworks to redefine the essence of LAWS-induced responsibility gaps as a phenomenon of "Concealment of Volition." The exercise of lethal state power is transmuted into the technologically neutral discourse of algorithmic computation, displacing the fundamental question of accountability—"who decided to kill" —into an invisible domain. Existing legal frameworks are rendered ineffective before this mechanism of power concealment. International Humanitarian Law (IHL)'s requirement for human ethical judgment is supplanted by probabilistic calculations, while national policies of the United States, United Kingdom, EU, and China render responsibility invisible through distinct mechanisms of abstraction, contextualization, exemption, and fusion. Domestic law similarly collapses through defects in command responsibility elements under military criminal law, structural imbalances in burden of proof under the State Compensation Act, and violations of the principle of statutory reservation through administrative rule-based regulation. Based on this diagnosis, the study establishes "Visibilization of Volition" as the fundamental principle of public law solutions. Visibilization of form mandates enactment of the "Act on the Regulation of Lethal Autonomous Weapon Systems," instituting a prohibition on fully autonomous weapons and requiring prior parliamentary consent for high-risk AI weapon systems. Visibilization of agency transforms MHC into legal obligations, demanding explanation of decision rationales, technologically enforced minimum deliberation time, and physical attack termination authority (Kill Switch). Visibilization of responsibility reverses the burden of proof to the state and establishes a dual responsibility structure distinguishing systemic supervisory negligence from operational procedural violations. Application of the proposed legal framework to the 2023 Israeli military Lavender system case validated its capacity for multi-layered regulation through democratic control at the adoption stage, MHC violation determination at the operational stage, and state compensation liability at the incident stage. This research contributes by reconceptualizing LAWS issues through power theory to demonstrate that responsibility gaps constitute not technological limitations but deliberate power concealment, while proposing fundamental directions for legislation. The fundamental principle of rule of law must be reaffirmed: decisions over life take precedence over efficiency, and lethal force must be attributed to accountable human agents.