인공지능에 대한 제조물책임법의 개선제안 - 유럽연합의 개정 제조물책임지침을 참고하여 -

초록

The European Union has revised its Product Liability Directive to bring software—the foundation of artificial intelligence—within the material scope of liability, and to include distributors within the “economic operator” concept, imposing joint liability. It also elaborates defect-assessment criteria into nine factors and introduces “disclosure of evidence" to ease victims' burden of proof. However, limiting standing to natural persons and excluding professional data damage leaves entities and professionals without relief, and treats unrealistic factors—“safety expectations,” uncontrollable “product influence,” and unattainable “cybersecurity”—as defect criteria. This article proposes Korean reforms drawing on the EU's Directive: adopt the expanded scope, recalibrate unrealistic criteria, and shift proof of defect to manufacturers. This accords with strict liability and surpasses the EU's disclosure regime in speed, cost, and secrecy protection. Appraisers should use “explainable AI” so victims can rebut manufacturers' defenses. For the “state-of-the-art defense,” manufacturers should manage algorithms, monitor guardrails, and disclose incidents. The “compliance” defense should exempt manufacturers forgoing compliance for safety; the “supplier's defense” abolish. •Key Words Revised Product Liability Directive, shifting the burden of proof, AI monitoring history, limitation on grounds for defense, mental health harm

키워드