重庆社会科学

生成式人工智能颠覆性变革知识内容生产方式并成为驱动千行百业创新发展的关键力量，探讨其数据训练过程的版权问题，可为人工智能规范发展提供政策借鉴。从版权侵权表征、版权保护对象及国际通行的豁免规则等方面深入探究生成式人工智能版权保护与豁免的辩证关系。研究发现，版权侵权问题可能遍历生成式人工智能运作的各阶段，应加强对传统资源数字化和原创数字内容的版权保护，务实的做法是通过豁免规则在发展中落实保护。并且，基于问题导向剖析欧盟、英国、日本的版权法律豁免规则，考察国际实践经验发现，各国的具体做法是设立“TDM例外”规则，中国可借鉴相关做法，在现行法律框架下嵌入基于特定行为目的的例外规则，从而兼顾版权保护与创新发展。

Generative artificial intelligence has disrupted the way knowledge content is produced and has become a key force driving innovation and development in various industries. Exploring the copyright issues in its data training process can provide policy references for the standardized development of artificial intelligence. This article delves into the dialectical relationship between copyright protection and exemption for generative artificial intelligence from aspects such as the manifestations of copyright infringement, the objects of copyright protection, and the internationally accepted exemption rules. The research finds that copyright infringement issues may occur at all stages of the operation of generative artificial intelligence. Therefore, it is necessary to strengthen the copyright protection of traditional resources digitization and original digital content. A practical approach is to implement protection through exemption rules during development. Moreover, based on problem-oriented analysis, this article examines the copyright legal exemption rules of the European Union, the United Kingdom, and Japan, and finds that the specific practices of these countries are to establish "TDM exceptions". China can draw on these practices and embed exception rules based on specific behavioral purposes within the current legal framework to balance copyright protection and innovation development.