摘要
生成式人工智能创作需要基于宏大体量数据的大模型训练,有着自由使用各类作品的现实需求,需要版权例外规则的制度保障。如今生成式人工智能创作的版权例外规则只有创作整体过程适用“合理使用”免责情形的唯一路径。然而生成式人工智能创作的版权例外规则还存在人工智能生成内容可能存在非版权性使用以及非法律意义使用的无责情形,且应当将生成式人工智能创作中的训练学习与分析表达两阶段分开分别进行法律评价。经过分析,应当构建生成式人工智能创作版权例外规则的双阶实现路径,即一是扩张合理使用条款,将机器学习、改写渲染的高度相似输出纳入其中;二是设置非法律意义使用的无责条款,将一般情况下的输出纳入其中。
The creation process of generative artificial intelligence relies on large-scale model training with massive datasets,necessitating unrestricted access to various works for which copyright exception rules must provide institutional safeguards.Currently,the only available copyright exception for generative AI creations is the application of"fair use"defense throughout the entire creative process.However,copyright exceptions for generative AI creations should also account for two additional scenarios:non-copyrightable uses of AI-generated content and non-legal significant uses.Furthermore,the generative AI creation process should be legally evaluated by separately examining its two distinct phases:the training/learning phase and the expression/analysis phase.Through comprehensive analysis,this study proposes a dual-path framework for implementing copyright exception rules for generative AI creations:Expanding fair use provisions to encompass machine learning processes and highly similar derivative outputs;Establishing a non-liability clause for non-legal significant uses to cover ordinary output circumstances.
作者
陈熠凡
林少东
Chen Yifan;Lin Shaodong
出处
《北京政法职业学院学报》
2025年第2期65-70,共6页
Journal of Beijing College of Politics and Law
基金
研究阐释党的十九届五中全会精神国家社会科学基金重大项目“在法治轨道上促进平台经济、共享经济健康发展研究”的阶段性研究成果,项目编号:21ZDA025。
