中圖分類號：D923；D922.17文獻(xiàn)標(biāo)識碼：ADOI:10.19358/j.issn.2097-1788.2025.12.011引用格式：徐榕嶺. 人工智能侵權(quán)立法中的歸責(zé)邏輯沖突與矯正［J］.網(wǎng)絡(luò)安全與數(shù)據(jù)治理，2025，44（12）：75-82.

Conflicts and corrections in the attribution logic of tort liability legislation for artificial intelligence

Abstract： The development of artificial intelligence (AI) technology poses significant challenges to tort liability legislation, particularly through the conflict between traditional attribution logic and the emerging risktiered liability framework. Theoretically, the risktiered attribution logic originates from the European Union′s riskbased governance scheme for AI. However, it disrupts existing civil tort liability structures, leading to uncertainty in liability allocation among stakeholders. By prioritizing technological risks over distinctions in protected legal interests, this logic inherently conflates the jurisdictional boundaries between administrative law and tort law. AIrelated risks—characterized by severe damages, high likelihood of evading litigation, and substantial informational costs—are difficult to regulate effectively under conventional rightsprotection rules. Consequently, it is imperative to revert to an attribution logic grounded in distinctions among legal interests. Concretely, a clear division of responsibilities between administrative and tort law must be established, with "public interest" serving as the demarcation criterion. When necessary, principled legal provisions may be adopted to address emerging risks in a limited manner, thereby achieving efficient governance of technological risks with minimal institutional costs.

Key words : artificial intelligence; riskbased regulation; tort liability legislation; civiladministrative interface