【作者】 马飞；

【导师】 夏静宜； 殷善鹏；

【作者基本信息】 扬州大学 ， 法律硕士（专业学位）， 2025， 硕士

【摘要】 生成式人工智能技术问世不仅推动了社会生产力的快速发展,也带来了更加复杂的人工智能生成内容的侵权责任认定问题。而通过“技术—法律”双重视角,可以洞悉在生成式人工智能时代人工智能生成内容侵权责任认定中的核心困境,在技术层面上,生成式人工智能依赖海量数据进行训练,并依靠自主生成能力生成内容,使得侵权行为更加不可控化和规模化。在法律层面上,现行法律框架在应对人工智能生成内容侵权责任认定问题时在侵权责任主体认定、因果关系证明上捉襟见肘,暴露出了现行法律规定的不适配性。为此应细化现行法规则,首先,对于侵权责任法的规定而言,应明确生成式AI“服务提供者”的核心责任主体地位,依据生成式AI的不同阶段构建分层归责体系,简化认定因果关系的标准,采用“事实上的关联性推定”的认定标准,细化技术中立原则的应用规定。对于著作权法规定应完善现行法规则,建立人工智能生成内容独创性的认定细化标准,统一司法实践判案规则;细化“实质性相似”判定标准,优化举证责任分配;并对《著作权法》第24条进行扩大解释,将符合规定的AI训练数据纳入"合理使用"的范围。对于人格权规定应重构知情-同意规则,对生成式人工智能训练所需数据作分层处理,以“不侵害他人人格权”作为合理使用的底线。其次,应沿着我国目前分散式人工智能立法思路稳步推进,在积攒足够的实践经验后再逐步建构综合性的人工智能法案。最后,构建针对生成式人工智能算法的治理框架也尤为重要,人工智能算法治理应当以对算法的监管、审查、责任分配为核心,构建贯穿事前审查、事中监管和事后追责救济三个维度的算法治理体系。即通过探究人工智能算法的技术性特征与法律规则的冲突本质,结合域外人工智能生成内容侵权治理的实践经验,提出具备理论创新和实践可行性的本土化治理路径。更多还原

【Abstract】 The advent of generative artificial intelligence(AI)technology has not only accelerated the rapid development of social productivity but also introduced more complex challenges in determining infringement liability for AI-generated content.Through the dual perspectives of"technology-law,"we can discern the core dilemmas in determining infringement liability for AI-generated content in the generative AI era.At the technological level,generative AI relies on massive datasets for training and autonomous content generation capabilities,rendering infringement activities increasingly uncontrollable and scalable.Legally,the existing legal framework struggles to address issues such as identifying liable entities and establishing causal relationships in AI-generated content infringement cases,revealing the inadequacy of current regulations.To address this,existing legal rules should be refined.First,regarding tort liability law:Clarify the"service provider"of generative AI as the core liable entity.Establish a tiered liability system based on different stages of generative AI operations.Simplify causality determination standards by adopting a"presumption of factual relevance"approach.Specify application guidelines for the principle of technological neutrality.For copyright law:Develop refined criteria for determining the originality of AI-generated content and unify judicial practice standards.Detail"substantial similarity"assessment standards and optimize burden of proof allocation.Expand the interpretation of Article 24 of the Copyright Law to include compliant AI training data under"fair use."Regarding personality rights:Restructure the informed-consent framework with stratified processing of training data.Establish"no infringement of personality rights"as the baseline for fair use.Second,China should continue its current decentralized AI legislative approach while accumulating practical experience,gradually progressing toward comprehensive AI legislation.Finally,constructing a governance framework for generative AI algorithms is crucial.Such governance should focus on algorithm supervision,review,and liability allocation,establishing a three-dimensional system encompassing pre-implementation review,real-time supervision,and post-event accountability.By analyzing the inherent conflicts between AI algorithm characteristics and legal norms,combined with international experiences in AI-generated content infringement governance,we can propose theoretically innovative and practically feasible localized governance pathways.更多还原