휴먼인더루프(human-in-the-loop) 시스템과 형사책임

초록

The Human-in-the-Loop (HITL) principle requires meaningful human oversight and intervention across the design, deployment, and operation of artificial intelligence (AI) systems. As AI systems attain higher levels of autonomy, the so-called “responsibility gap”—in which legal and moral accountability for harmful outcomes becomes diffuse or indeterminate—tends to widen. The HITL principle seeks to mitigate this gap by securing effective and continuous human control throughout the AI decision-making processes. Within the HITL framework, individuals involved in the development, deployment, or operation of AI systems that cause harm should, in principle, bear criminal liability commensurate with the nature and extent of their involvement. The central difficulty, however, lies in delineating the proper scope and limits of such liability. Because HITL systems are inherently structured around human–AI collaboration, the attribution of criminal responsibility must account for both the form and degree of human intervention and the operational characteristics, including the level of autonomy, of the AI system. In this respect, there is a compelling need for an intermediate sanctioning model capable of addressing behavior that gives rise to risks falling between traditional criminal law and administrative regulation. A particularly promising approach is a mitigated model of criminal liability inspired by Winfried Hassemer’s concept of Interventionsrecht (Intervention Law), which offers a normative framework for responding to risks that cannot be adequately managed through either conventional criminal sanctions or administrative measures alone.

키워드