Research on Copyright Infringement Issues of Generative Artificial Intelligence
Lei Jiang · Law and Humanities · 2025
This article takes two AI infringement cases involving Ultraman in Hangzhou and Guangzhou as entry points to analyze the infringement disputes of gene-ra-tive artificial intelligence in the stages of data input, model training, and con-tent output. In the Hangzhou case, the platform was found to be liable for con-tributory infringement because users generated infringing content using its tools, and the court recognized fair use during the training stage. In the Guang-zhou case, the platform directly generated infringing content, was de-termined to be directly infringing, and fair use was not applicable. The differ-ences in the judgments of the two places reflect the liability stratification of "technology tool providers" and "content service providers", as well as the bal-ance struck by the judiciary between technological innovation and copyright protection. It is recommended to add special provisions on fair use of AI train-ing, establish flex-ible judgment criteria such as the "three-step test" and "transformative use", clarify the liability boundaries of different service models, and promote the co-ordinated development of technological innovation and rights protection.