인공지능은 계약의‘도구’인가, 아니면‘대리인’인가? - AI가 제기하는 계약법적 쟁점에 대한 소고(小考) -

Artificial Intelligence (AI) has now evolved into ‘Agentic AI (AAI)’, a stage where it can autonomously perform complex tasks by automatically linking with other AIs or applications with minimal human intervention. Premised on AAI as the current state-of-the-art technology, this article seeks to investigate, through legal interpretation, the core questions in contract law posed by AI: ‘contract formation’ and ‘risk allocation’. For contract formation, this article argues for the validity of the ‘Tool Theory,’ which perceives AI as the user’s tool. The ‘Agency Theory’ is presented as an unduly broad legal fiction, as it problematically confers agency status upon a non-entity lacking legal personality. In contrast, the Tool Theory aligns with the principle of self-responsibility, is applicable regardless of technological advancement, and is consistent with international norms like the UNCITRAL Model Law on Automated Contracting. Concerning risk allocation, this article examines the doctrine of mistake as a coherent extension of the Tool Theory, preferable to the doctrine of apparent agency in terms of theoretical consistency and the rational allocation of the burden of proof. Although a user's attempt to rescind a contract on such grounds is unlikely to succeed—as it may be construed as an 'assumption of risk' or a mere 'mistake in motive'—a practical safety net exists. Most AI-based transactions are consumer contracts, where users can be protected by exercising their statutory right of withdrawal under the Electronic Commerce Act. As AI fundamentally reshapes society, the field of law must act not as a passive follower of technology but as a proactive partner, understanding the principles of change to guide legal regulation.