| Court | Hangzhou Intermediate People's Court, Zhejiang province |
| Country | Mainland China |
| Parties | Claimant: Shanghai Xinchuanghua Cultural Development Co., Ltd. Defendant: Hangzhou Shuimu Intelligent Technology Co., Ltd. |
| Date Case Filed | 20 February 2024 (First Instance); 28 October 2024 (Second Instance) |
| Type of Claim | Copyright infringement and unfair competition |
| Date Judgement issued | 25 September 2024 (First Instance); 30 December 2024 (Second Instance) |
| Summary of Key Background Facts | The claimant holds the copyright in an animated character. The defendant operates an AI platform that allows users to upload the claimant’s copyrighted works to train an AI model and generate new images. The claimant alleged that the defendant, as an AI service provider, facilitated the creation of infringing content through its training data, thereby infringing the claimant’s copyright and constituting unfair competition. |
| Remedies sought | The claimant sought the following remedies: - Cease all infringing acts, including suspending services for training and publishing related models, deleting all related training models, images, materials and data on the platform, as well as shielding all unauthorised use of the animated character and its associated keywords across the entire platform;
- Compensate the claimant for economic losses and reasonable legal expenses totalling RMB 300,000.
|
| Summary of key legal arguments | Claimant: The defendant operated an AI platform that permitted users to upload the claimant’s copyrighted animated character works to train AI models and generate similar derivative images. The claimant alleged that the defendant, as the platform operator, failed to block such unlicensed training and dissemination, facilitated mass infringing content production, and thus committed contributory copyright infringement as well as unfair competition. Defendant: - The platform merely provides neutral generative AI technical tools including AI model training and image generation functions, and all training materials involving the claimant’s animated character works are independently uploaded by users rather than supplied or guided by the defendant.
- The defendant has inserted IP disclaimer clauses in user agreements and launched a notice-and-takedown mechanism to handle potential infringing content, having fulfilled the basic platform obligations under safe harbour rules.
- The platform’s AI training service itself has extensive legitimate non-infringing application scenarios, so its neutral technical business model shall not be deemed as intentionally facilitating copyright infringement by users.
- The same act shall not be subject to dual evaluation under copyright law and anti-unfair competition law.
Court’s Judgement: - The platform failed to implement adequate measures to prevent copyright infringement of its users, which constitutes contributory copyright infringement. Therefore, the defendant is ordered to immediately cease the acts of infringing the claimant’s right of communication via information network and to compensate RMB 30,000 to the claimant for economic losses and reasonable expenses.
- The development of AI requires the input of massive volumes of training data, which inevitably involves the use of third-party copyrighted works. Generative AI typically comprises four key stages: data input, model training, content generation and content utilization. From the perspective of fostering the advancement of generative AI technologies, courts favour a more lenient standard for assessing potential infringement in the data input and model training phases. Conversely, a stricter standard is deemed appropriate for evaluating infringement related to the output and subsequent use of AI-generated content. This differentiated approach aims to reach a balance between promoting technological innovation and protecting intellectual property rights.
|
| CMS Comment | This case marks the first judgment in China that data training for generative AI models complies with fair use under the PRC Copyright Law. It constitutes an important breakthrough for fair use rules in artificial intelligence training. |