生成式人工智能侵权责任探析

Tort Liability of Generative Artificial Intelligence

生成式人工智能的侵权责任,在理论上可以从产品责任和一般侵权责任两种路径进行分析。在当前的法律框架下,对于以大语言模型为代表的生成式人工智能系统而言,不宜将其纳入到产品责任的范畴,而宜通过一般侵权责任制度并辅之以过错推定规则,这样既可以解决受害人的举证困难问题,也便于通过司法控制机制对生成式人工智能提供者的责任负担进行动态调整。为了维持生成式人工智能提供者与受害人之间的利益平衡,可以通过“通知—处置”规则对提供者施加消除侵权信息影响并防止系统再次生成侵权信息的义务。

The liability for infringement caused by generative artificial intelligence(AI)can be analyzed theoretically from two paths:product liability and general tort liability.However,within the current legal framework,it is not appropriate to classify generative AI systems,characterized by large language models,under product liability.Instead,it is more suitable to address them through general tort liability,accompanied by rules of presumed fault.This approach not only helps resolve the burden of proof faced by the victims,but also enables the dynamic adjustment of liability for generative AI providers through judicial control mechanisms.To maintain a balance of interests between generative AI providers and the victims,a"notice-and-takedown"rule can be implemented,imposing an obligation on providers to eliminate the impact of infringing information and prevent the system from generating such information again.