摘要
生成式人工智能具有自主性、不透明性、不可预测性等特征,这也导致了学界认为人工智能侵权归责需脱离产品责任,理由主要为产品责任为严格责任,生产者难以控制人工智能运行过程中存在的风险,要求其承担严格责任并不公平;生成式人工智能大多以服务的形态存在,因此其不符合产品的定义。然而,从法律属性来看,生成式人工智能提供者与网络服务提供者二者存在实质差异。生成式人工智能系统符合产品的定义,产品责任契合生成式人工智能系统多层次、多样态的特征,能够实现与监管的良性互动。在产品责任框架下,开发者与运营者可以类比为产品责任中生产者与销售者的地位,可以根据生成式人工智能的生命周期、层次属性确定制造缺陷、设计缺陷、警示缺陷、跟踪观察缺陷,确定发展风险抗辩适用的条件与标准。
Two primary pathways exist for addressing liability in cases of generative Al infringement.Much of the existing academic discourse centers on the liability of network service providers,with less fo-cus on product liability-which some argue should be categorically excluded as strict liability.Generative Al is frequently classified as a network service provider because it typically lacks standardized,patterned product features.Under this classification,fault-based liability applies,allowing policy considerations to influence assessments of fault.However,from a legal perspective,generative AI providers fundamentally differ from traditional network service providers.Generative AI more closely resembles a tangible prod-uct,and liability under product law better captures its multi-layered,multi-state nature-thus facilitating regulatory oversight.Additionally,applying product liability principles requires that standards for devel-opment risk defenses and defect assessments-such as manufacturing,design,warning,and tracking de-fects-take into account generative AI's life cycle and hierarchical complexities.
出处
《比较法研究》
北大核心
2025年第3期69-84,共16页
Journal of Comparative Law
关键词
生成式人工智能
网络服务提供者
产品责任
generative artificial intelligence
network service providers
product liability
作者简介
李雅男,华中科技大学法学院讲师,法学博士。
