인공지능과 형법 책임 귀속의 한계와 가능성

Abstract

The advancement of artificial intelligence (AI) in the modern information society has introduced new legal and ethical challenges. While AI offers significant convenience, it simultaneously presents risks that pose a challenge to the existing criminal law framework. This paper aims to explore the legal response of criminal law to infringements on legal interests caused by AI, as well as the attribution of responsibility. In the current information society, the rapid development of information technology has led to emerging forms of crime and violations of legal interests, exemplified by cases such as hacking, data breaches, and deepfake-related sexual offenses. The information society demands a re-evaluation of individual and collective legal interests, necessitating changes in the role and function of criminal law. As technology advances, society faces potential risks and self-destructive possibilities, making preventive measures within criminal law increasingly important. AI technology operates based on algorithms, differing fundamentally from human actions driven by autonomy, which raises complex questions about how to assign responsibility. Although AI can independently make judgments and take actions, clear criteria for attributing criminal responsibility for its outcomes remain undeveloped. The traditional criminal law system, which relies on human conduct and accountability, is encountering disruptions due to the rise of AI. To address the issue of attributing criminal responsibility to AI, approaches such as ‘dehumanizing corporate personality’, ‘reconstructing the concept of the legal person’, and ‘analogy-based reasoning’ have been proposed. However, these methods face inherent limitations, as AI lacks free will and the capacity for moral judgment, making it difficult to recognize AI as a subject under criminal law. AI presents risks characterized by autonomy, unpredictability, and opacity, which complicates the application of traditional criminal liability frameworks. Therefore, applying the principle of “permitted risk” is suggested, allowing AI operators to avoid liability for socially acceptable levels of risk rather than being held accountable for all potential risks. This approach supports technological innovation by reasonably limiting the scope of criminal liability without hindering progress. To prevent infringements on legal interests caused by AI, it is crucial to impose an enhanced duty of care on AI operators. This paper identifies specific duties of care, including the obligation to manage data quality, ensure safety, and conduct thorough testing and verification. Given the evolving risks associated with AI, the criminal law framework must adapt accordingly. While directly imposing criminal liability on AI remains impractical at present, responsibility should be attributed to the natural or legal persons who design and manage AI systems. Additionally, criminal law must adopt a preventive role by extending negligence liability principles and clearly defining the duty of care, ensuring a balanced management of both the benefits and risks that AI technology introduces.