生成式AI数据使用中的著作权侵权风险研究
Research on Copyright Infringement Risks in Generative AI Data Use

生成式AI展现出了无限潜力，但同时也引发出一系列问题，无论是在模型输入阶段数据使用的侵权，还是模型输出阶段能否构成作品，这些问题都引发当前学界极大的争论。本文聚焦生成式AI输入阶段数据使用的著作权侵权问题，明确侵犯著作权的具体化表现，解构所侵犯的复制权和改编权类型，借鉴欧盟、美国等国家的经验，并提出解决侵权问题需将生成式AI的数据使用纳入法定许可范围并降低使用费用，或引入转换性使用制度对合理使用进行扩展，同时在AI发展的整个过程中呼吁国家注重监管，只有如此才能更好的寻求生成式AI未来的发展之道。
Generative AI has shown infinite potential, but at the same time, it has also raised a series of issues, whether it is infringement of data usage in the model input stage or whether the model output stage can constitute a work, these issues have sparked great debate in the current academic community. This article focuses on the copyright infringement issue of data usage in the input stage of generative AI, clarifies the specific manifestations of copyright infringement, deconstructs the types of replication and adaptation rights infringed, draws on the experience of countries such as the European Union and the United States, and proposes that solving the infringement problem requires incorporating the use of generative AI data into the legal licensing scope and reducing usage costs, or introducing a transformative usage system to expand fair use. At the same time, it calls on the state to pay attention to regulation throughout the entire process of AI development, in order to better seek the future development path of generative AI.