China’s first AI-generated image copyright infringement case

When someone shared an image made using StableDiffusion on social media, the ensuing case drew wide attention and set a precedent, explains Hongxia Wu of CCPIT Patent and Trademark Law Office.

In November 2023, the Beijing Internet Court issued the first-instance judgment on what is considered China's first copyright infringement case involving AI-generated images.

The case has received significant attention from society, with 170,000 netizens tuning in to the online live broadcast of the trial. The court ruled that the AI-generated image in question is entitled to copyright protection. This ruling, which is now in effect, has far-reaching implications for the judicial safeguarding of copyright in AI-generated images and the advancement of the AI industry in China.

Case overview

A Chinese individual, the plaintiff, created an image using the StableDiffusion open-source software on February 24, 2023, and shared it on a social platform. Subsequently, another individual, the defendant, used the plaintiff's image in an article posted on March 2, 2023. The plaintiff filed a lawsuit, alleging infringement of his authorship and information network dissemination rights.

The Beijing Internet Court determined that the image constituted intellectual work in literature, art, and science, possessing originality. The court recognised the plaintiff as the rightful copyright holder of the image and found the defendant guilty of infringing on the plaintiff's rights, awarding 500 yuan ($69) in compensation for economic losses.

Societal controversies

The case sparked widespread debates regarding the relation between AI-generated content and copyright in China. Opinions varied, with some supporting and others refuting the copyrightability of AI-generated images. Notable figures, including law professors, lawyers, and researchers, shared their perspectives on the matter.

The primary argument against the copyrightability of AI-generated content lies in the level of control and influence that AI users have over the resulting content.

While users provide prompts to AI and request it to complete a specific task, they are unable to entirely determine the expressive elements of AI-generated content according to their wills.

Essentially, due to the inherent nature of "autonomous learning" within artificial intelligence, the creation of its content surpasses the user's direct control.

Court's verdict

In light of the case's significance, the Beijing Internet Court thoroughly examined three core issues in its judgment: the plaintiff's intellectual investment, the originality of the image, and the ownership of copyright.

1. The court affirmed that the image reflected the plaintiff's intellectual efforts in directing the AI model to generate the image through prompts and parameter settings.
2. It found that the image has originality as the plaintiff actively shaped visual elements and made aesthetic choices during the creation process.
3. The court rejected that the AI model or its designer could claim authorship, instead affirming the plaintiff's ownership of copyright upon the image due to his direct involvement in generating and selecting the image.

Value-based judgment

The court's decision not only addressed legal aspects but also underscored the value of encouraging creative pursuits. The evolution of AI technology has brought about significant changes in the way humans create.

As highlighted in the court's verdict, advancements in technology and the increasing intelligence of tools may reduce the extent of human input in the creative process.

However, this does not diminish the importance of applying the copyright system to promote the creation of works. By recognising AI-generated content as works and AI users as authors, the judgment aimed to incentivise the creation of works using new tools and foster innovation in the AI industry.

Impact on copyright

The ruling would set a precedent for future cases involving AI-generated images, signalling a likely trend of recognising AI-generated images as works and attributing copyright ownership to users rather than AI models.

However, the determination of whether specific AI-generated images qualify as works will depend on the specifics of each case. In general, when individuals utilise artificial intelligence for image generation, the diversity of their requests, the clarity and specificity of their descriptions regarding image elements, layout, and composition, and the level of personalisation in their expression all contribute to the originality of the generated images. Conversely, images created with a single click through artificial intelligence are less likely to be acknowledged as original works.

Recommendations on utilising AI in China:

1. To secure copyright protection for AI-generated images, it is crucial to provide detailed and distinct descriptions of image elements and layout compositions to AI. By emphasising unique characteristics and personalisation, AI users can meet the originality criteria required for copyright protection.
2. Maintain a record of the AI image generation process, including the selection of AI models, input prompts, reverse prompts, and parameter adjustments, to establish a basis for proving personalised expression in the generated images, as the plaintiff did in this case.
3. When engaging in design outsourcing, ensure that copyright ownership by you upon AI-generated content is clearly defined in the agreements between you and design companies, as well as between design companies and their employees.

By following these guidelines, you can navigate the evolving landscape of AI technology while safeguarding creative outputs and legal interests in China.

Hongxia Wu is a trademark and copyright Litigation attorney in China and deputy director of Trademark & Copyright Litigation at CCPIT Patent and Trademark Law Office. She can be contacted at wuhx@ccpit-patent.com.cn