摘要
生成式人工智能在模型训练过程中对海量作品数据的使用,面临著作权法上复制权侵权的认定困境。当前法律对“复制”行为的宽泛界定,未能区分技术性复制与具有传播目的的复制,导致数据处理合法性高度不确定。研究提出,应在司法实践中对复制权进行目的性限缩,以“固定性”与“传播性”作为构成侵权的双重标准。训练过程中所涉临时性、非表达性复制行为,因不具备独立传播意图与市场替代效果,不应落入复制权规制范围。在此基础上,数据溯源技术可作为验证“非传播目的”与过程合规的关键治理工具,通过记录训练全流程元数据,构建可信证据链,为司法裁判提供客观依据。最终,应构建法律解释与技术治理相协同的分层实施框架,在不妨碍技术创新的前提下实现著作权制度的有效调适。
The use of massive copyright data in the training of generative AI models raises significant challenges in determining copyright infringement under the right of reproduction.The broad legal definition of“copying”fails to distinguish between technical replication and reproduction for dissemination purposes,creating substantial uncertainty regarding data processing legality.This study advocates for a teleological restriction of the reproduction right in judicial practice,proposing“fixity”and“disseminability”as dual criteria for infringement.Temporary and non-expressive copying during training,which lacks independent dissemination intent and market substitution effect,should fall out of the scope of the reproduction right.Furthermore,data traceability technologies serve as crucial governance tools for verifying“non-disseminative purpose”and process compliance,by recording entire training metadata to build credible evidence chains for judicial reference.Ultimately,a layered governance framework integrating legal interpretation and technological measures should be es-tablished to effectively adapt the copyright system without impeding technological innovation.
作者
王若宇
胡神松
WANG Ruoyu;HU Shensong(School of Law and Humanities&Socidogy,Wuhan University of Technology,Wuhan 430070,Hubei,China)
出处
《昆明理工大学学报(社会科学版)》
2025年第6期32-40,共9页
Journal of Kunming University of Science and Technology(Social Sciences)
基金
国家社会科学基金项目“新媒体时代舞蹈作品版权多维保护机制的构建与完善研究”(18BE077)
中国法学会部级课题“基于风险动态评估的脑机接口技术分级监管机制研究”[CLS(2025)C24]
武汉市法学会重点课题“网络直播风险多元化治理研究”(2023006)。
关键词
生成式人工智能
数据溯源
复制权
技术治理
非表达性使用
generative AI
data traceability
right of reproduction
technological governance
non-expressive use
