China: Documentation of creative processes is crucial to claim copyright in AI-generated works

China: Documentation of creative processes is crucial to claim copyright in AI-generated works 

A recent decision by the Beijing Internet Court represents a significant development in Chinese copyright law’s approach to content generated using AI tools. This landmark judgment provides guidance on evidentiary standards for creators seeking copyright protection in AI-generated works in mainland China.

Copyright subsistence in mainland China

Under Article 3 of the PRC Copyright Law, works are defined as intellectual expressions in the fields of literature, art and science that possess originality, and that can be reduced into a certain tangible form. The emergence of AI as a creative tool has prompted the Chinese courts to respond proactively to the novel legal challenges it presents, particularly with respect to the requirement of originality in copyright works. In 2024, the Guangzhou Internet Court delivered mainland China’s first judgment on copyright infringement involving AI-generated content, finding the provider of a generative AI service liable in the “Ultraman” case. This landmark decision was quickly followed by a ruling from the Beijing Internet Court, which addressed mainland China's first case concerning the infringement of personality rights arising from an AI-generated voice. These decisions established the foundation for courts to address AI-generated content, setting the stage for this latest ruling on AI-generated images.

Background of decision

This latest 2025 decision in relation to AI-generated images concerns a content creator (“Plaintiff”), who sought to assert copyright in an image generated using an AI tool during a business partnership with a Beijing technology company (“Defendant”). The Plaintiff discovered that the Defendant had used the image without permission for promotional activities across multiple platforms, and claimed copyright infringement against the Defendant.

The Defendant contested the Plaintiff’s copyright ownership on several grounds. Firstly, the Defendant argued that the image was not the Plaintiff’s original creation, as both parties had collaborated on selection of material and prompt creation. The Defendant also claimed that the Plaintiff failed to prove the requisite threshold of originality, and further denied any commercial use in its use of the image.

The case hinged on a critical evidentiary gap, as the Plaintiff did not provide original generation process records from the AI tool, instead offering only attempted post-hoc recreations during the litigation.

Judgment

The Beijing Internet Court’s decision centred on whether the AI-generated image constituted a “work” under the PRC Copyright Law, with particular focus on the originality requirement. 

The court emphasised that determining “originality” of AI-generated content follows the same fundamental principle of placing the burden of proof on the asserting party as in the case of other copyright works. In other words, the party claiming copyright in AI-generated works bears the burden to prove that it has invested creative process in the generation of the image.

Crucially, the court ruled that the party asserting copyright ownership in AI-generated content must explain its creative thinking, input commands, and the process of selecting and modifying generated content, supported by relevant evidence. The court noted that, whilst AI generation involves less direct human intellectual input compared to conventional tools like paint brushes or cameras, this does not preclude copyright protection if it can be demonstrated that sufficient creative effort has been expended. 

In the present case, the court found that the Plaintiff’s evidence fell short of the required threshold, due to the absence of original generation process records, actual prompts or generation commands, or documentation of the creative choices. Therefore, the image was not subject to copyright protection under the PRC Copyright Law. While the Plaintiff had attempted simulations of the generation process during litigation, the court rejected these post-hoc recreations as evidence, on the basis of them being inadequate and lacking consistency and comparability with the original generation process.

Commentary

This final judgment reflects the court’s recognition that AI tools remain instruments that require human direction and creative input to produce copyrightable works.

The court’s emphasis on “documentation of creative processes” reflects the unique challenges posed by the use of AI. Unlike traditional creative processes where human contribution is more readily apparent, AI generation requires more detailed evidence to demonstrate the creator’s intellectual input. This approach aligns with the fundamental principle that copyright requires human authorship.

The decision’s practical implications extend beyond individual creators to the broader AI industry. The Beijing Internet Court’s recommendation for enhanced traceability capabilities suggests that AI platform developers should implement features that automatically preserve generation logs, prompts, and iterative processes. For businesses and creators operating in mainland China, this underscores the critical importance of implementing robust documentation protocols when using AI tools, ensuring that evidence of human creative input is preserved from the outset of the creative process rather than attempting reconstructions that courts will likely reject.