We previously wrote about a Chinese court decision that protected the copyright of AI-generated images in the ‘half heart’ case. However, in a recent case, the Zhangjiagang People’s Court (in China’s Jiangsu province) denied copyright protection for AI-generated works.
In the new ‘butterfly chairs’ case, the plaintiff used the AI tool Midjourney to create three images. Each image depicted a distinct butterfly-shaped chair with a jelly-like texture in varying hues of pink, blue and orange.
The plaintiff posted the three images on the social media platform Xiaohongshu (RedNote), together with the prompts she entered: “Children’s chair with jelly texture, shape of cute pink butterfly, glass texture, light background”. In the post, the plaintiff sought to attract interest from furniture manufacturers in the designs. The first defendant approached the plaintiff on behalf of the second, third and fourth defendants, expressing interest in a licence to commercialise and mass-produce the butterfly chairs. However, the negotiations fell through. Despite this, the defendants proceeded to use Midjourney to generate similar designs by making references to the plaintiff’s publicly shared prompts, and then to, after multiple rounds of modification, manufacture and sell butterfly chairs based on the final design.
The plaintiff sued the defendants, demanding that they destroy their inventory of butterfly chairs, cease production and sales, and compensate the plaintiff RMB 200,000 for losses. The court ruled in favour of the defendants, dismissing the plaintiff’s claims on the grounds that the plaintiff’s AI-generated images were not protected by copyright and that the defendants were not liable for copyright infringement. The plaintiff unsuccessfully appealed to the Suzhou Intermediate People’s Court (in China’s Jiangsu province), rendering the 19 March 2025 decision by the Zhangjiagang People’s Court final and effective.
In reaching its decision, the court distinguished between types of AI-generated content. Content produced by a user exercising their original human intellectual input in the creation process with the assistance of AI as a tool is protected by copyright. Meanwhile, content mainly autogenerated by AI, with minimal human contribution, is not protected by copyright. Applying this to the butterfly chairs, the court found that the plaintiff’s three images fell into the latter category of AI-generated content, with insufficient original human intellectual input to be copyrightable. On the evidence, despite reusing almost identical prompts, the plaintiff was unable to regenerate the exact same images of the three original butterfly chairs using Midjourney. This highlighted the inherent randomness and unpredictability of AI-generated outputs, which was acknowledged by the plaintiff. Indeed, the significant variations between the regenerated versus original images of the butterfly chairs underscored the lack of true, author-driven expressions in the designs.
Importantly, the court outlined a test for determining whether AI-generated content includes sufficient original human intellectual input to qualify for copyright protection. The user must provide clear documentation recording the creation process of the AI-generated content, demonstrating that AI merely assisted as a tool rather than being the primary creator. In particular, there should be unique choices and modifications by the human user — such as refining prompts and adjusting input parameters — to alter the initial output generated by AI. Here, the plaintiff failed to provide sufficient evidence to meet this standard; thus, the court denied copyright protection for the plaintiff’s three butterfly chair images.
Our comments
The ‘butterfly chairs’ case is the third case to date in China addressing whether AI-generated images can be protected by copyright — a topical issue of growing global relevance. Unlike the first two cases, where the Chinese courts granted copyright protection to the AI-generated images, this third case appears to mark a departure, suggesting that the Chinese courts are diverging from their previous approach to copyright protection for AI-generated works.
However, a closer examination of the reasoning behind the outcomes of the three Chinese cases suggests that this is not the case. In fact, the fundamental principle, namely, that there needs to be sufficient original human intellectual input in the ultimate AI-generated work for copyright to subsist remained the same and has been applied uniformly across all three cases.
The contrasting outcomes are explained by the facts, which helpfully illustrate the circumstances where AI-generated works will be afforded copyright protection versus where they will not. The first two cases involved an iterative process whereby the plaintiffs refined prompts, adjusted parameters, generated images with Stable Diffusion or Midjourney, and further edited the outputs (e.g., using Photoshop). In contrast, the third case is distinguishable on its facts. The plaintiff sought copyright protection for the initial output of AI-generated images, created through a single pass of inputting prompts into Midjourney, without further modification. These fact patterns suggest a key distinction: AI-generated works created via iterative, human-guided processes are more likely to be protected by copyright, while AI-generated works created via a non-iterative, single-pass input process are less likely to be protected by copyright. As a rule of thumb, it seems that initial AI-generated outputs are highly unlikely to be protected by copyright under Chinese law.
The key takeaway is that Chinese courts have neither guaranteed copyright protection for all AI-generated works in the first two cases nor have they suddenly imposed a blanket restriction on copyright protection for AI-generated works in the latest case. Rather, the case reinforces the core purpose of copyright law in China: to safeguard human creativity.
Nevertheless, the fact that at least one Chinese court has denied copyright protection for AI-generated works suggests the willingness and potential for Chinese courts to decide similarly in future cases, aligning more closely to the United States’ practice. Comparing the numbers, the United States Copyright Office has refused copyright protection for AI-generated visual artworks in at least four cases. Guidance issued by the United States Copyright Office in March 2023 and January 2025 highlight that copyright only protects human-created works; copyright protection is not available for AI-generated content unless there is human contribution to AI-assisted works, assessed on a case-by-case basis; prompts alone insufficiently justify copyright protection for the resulting output. Thus, the ‘butterfly chairs’ case possibly signals a trend towards convergence between the United States and China on this issue, with both jurisdictions emphasising the necessity of human input for granting copyright protection.
[View source.]
