Managing Copyright Infringement Risks in Generative Artificial Intelligence Data Mining

In the era of artificial intelligence, the rapid development of generative artificial intelligence represented by ChatGPT brings great convenience to human creation, but also causes many potential copyright risks and brings a series of new challenges to the field of intellectual property. The creative process of generative artificial intelligence mainly includes four stages: training data input, data learning, input instruction and content generation. Among them, the legal use of copyright works in the data input stage needs to be solved instantly. In order to solve this problem, we should first improve the “fair use system” under current Copyright Law, divide machine learning into expressive and non-expressive types according to whether generative artificial intelligence has expressive content output, and discuss whether it constitutes fair use separately. Secondly, in order to protect the legitimate rights of original copyright owners, it is also necessary to improve the transparency of training data and increase the “opt-out” mechanism. In addition, it is also necessary to clarify the tort liability subject of generative artificial intelligence by legislation when it does not constitute fair use.