Mengchun’s Comments
This case provides a more structured approach to assessing the copyrightability of AI-generated content in China.
The court divided the creation process into three stages: pre-generation conception, generation control, and post-generation editing. It then developed a three-part test for assessing originality at the generation-control stage.
Compared with the earlier Zhangjiagang Court decision, this case takes the analysis a step further by providing a clearer framework for assessing how human input affects the specific expression of AI-generated content.
The present case involved only generation control and post-generation editing, and the court found neither sufficient to establish originality.
I. Key Issue
Under what circumstances can an AI-generated image satisfy the originality requirement for copyright protection under Chinese law?
II. Court’s Interpretation
The Changsha Kaifu District People’s Court held that whether an AI-generated image constitutes a copyrightable work should not be determined merely by the fact that artificial intelligence was used as a tool.
Instead, the court should examine whether the resulting image reflects original intellectual input by a natural person and otherwise satisfies the statutory requirements for a work under China’s Copyright Law.
The court divided the process of creating AI-generated content into three stages:
1. Pre-Generation Conception
Pre-generation conception refers to circumstances where, before using an AI system, the user has already developed sketches, compositional plans, or other expressive designs and instructs the AI system to generate content based on those designs.
The court noted that originality at this stage may be assessed in accordance with traditional principles applicable to human creation.
The present case, however, did not involve pre-generation conception. The court therefore did not conduct a specific originality assessment at this stage.
2. Generation Control
Generation control refers to the user’s control over the generation process through prompts, parameter settings, selection of results, and similar measures.
The court considered this stage materially different from traditional forms of creation and held that originality should be assessed by reference to fundamental copyright principles.
In particular, the court identified three factors:
- Whether the control directly concerns expression. Whether the prompts and parameter settings, beyond merely describing the subject matter, further concern specific expressive elements such as the appearance of the subject, composition, colors, and lighting, rather than only general requirements concerning subject matter or style.
- Whether there is a correspondence between the prompts and the resulting image. Whether the specific expressive characteristics of the final image can be matched with the user’s prompts and parameter settings.
- Whether the user’s choices dominate the expression of the image. Whether those corresponding characteristics cover the dominant expression of the image rather than merely background or subsidiary details.
3. Post-Generation Editing
Post-generation editing refers to substantive modifications made by the user after the AI-generated content has been produced.
The court noted that originality at this stage may also be assessed under traditional copyright principles.
The court emphasized that the three stages should be examined separately rather than conflated into a single assessment of originality.
III. Case Background
On May 8, 2025, a third party used the Jimeng AI image-generation model to create an image featuring five colored sachets associated with the Dragon Boat Festival.
The user entered the Chinese prompt equivalent to:
“Dragon Boat Festival sachets, colorful, Chinese style, with the Chinese characters ‘Dragon Boat Festival’.”
The user also selected a 1:1 image ratio and 1K resolution.
The AI system generated an image containing five sachets of different colors together with Chinese characters. The user subsequently edited the image in Photoshop, including removing two redundant Chinese characters generated in the lower portion of the image.
The image was first published on Xiaohongshu on May 17, 2025.
On June 3, 2025, the original user entered into an agreement transferring the economic copyright interests in the image to the plaintiff. The image was subsequently registered with the Liaoning Provincial Copyright Administration on June 18, 2025.
Separately, on May 27, 2025, the defendant, a kindergarten in Changsha, used the image in an article published through its WeChat official account concerning the Dragon Boat Festival holiday.
The plaintiff brought copyright infringement proceedings, alleging infringement of the right of communication through information networks and seeking CNY 10,000 in damages and reasonable expenses.
The defendant disputed, among other things, whether the AI-generated image possessed sufficient originality to qualify as a copyrightable work.
IV. Court’s Holding
In this case, the court found that the creation process involved only generation control and post-generation editing, but not pre-generation conception. It therefore examined originality at those two stages separately and held that neither reflected sufficient originality.
Generation Control Did Not Establish Originality
The court found that the prompts used to generate the image primarily concerned the subject matter, colors, and general style of the requested image.
Expressions such as “colorful” and “Chinese style” constituted general requirements concerning color and style. They did not make individualized choices regarding specific expressive elements such as the shape of the sachets, composition of the image, arrangement of elements, or lighting.
Similarly, the 1:1 image ratio and 1K resolution were ordinary technical parameters rather than expressive choices.
Accordingly, the court held that the user’s generation-control activities did not directly concern specific expression and therefore did not demonstrate originality.
Post-Generation Editing Did Not Establish Originality
The court compared the image before and after the Photoshop editing.
It found that the images were substantially the same, with the principal difference being the removal of redundant Chinese characters generated by the AI system.
The court characterized this as a technical correction of a defect, rather than a substantive modification of the image’s composition, colors, or visual effect.
The post-generation editing therefore did not create original expression sufficient for copyright protection.
The AI-Generated Image Was Not a Copyrightable Work
Because the image lacked sufficient originality at both the generation-control and post-generation editing stages, the court concluded that it did not constitute a “work” protected under China’s Copyright Law.
The plaintiff’s copyright infringement claims were therefore dismissed.
V. Judgment
On August 5, 2026, the Changsha Kaifu District People’s Court dismissed all of the plaintiff’s claims.
Court:
Changsha Kaifu District People’s Court
Decision Date:
August 5, 2026
Case No.:
(2026) Xiang 0105 Min Chu No. 8031
