Do AI generated images and articles shall be deemed as “works” protected by the “Copyright Law”?

At the end of November 2023, a judgment by the Beijing Internet Court sparked heated discussions. In this case, the plaintiff used AI to generate images and published them on Xiaohongshu (an app), the defendant used those images in his article. The plaintiff filed a lawsuit to the court on the grounds of copyright infringement. The first instance judgment of the Beijing Internet Court held that the images generated by AI in this case should be deemed as “works” protected by the “Copyright Law”, so the defendant has constituted a copyright infringement.

In this case, the Beijing Internet Court pointed out that, “The key to determining whether those images constitute ‘works’ lies in whether it is an intellectual achievement and whether it has originality.”.

Firstly, regarding the recognition of “intellectual achievements”, the court held that, “During the whole process from the conception section to the final selection section, the plaintiff made a certain amount of intellectual investment, such as designing the presentation of characters, selecting prompts, arranging the order of prompts, setting relevant parameters, selecting images which meet expectations, and so on. The images in the case reflect the plaintiff’s intellectual investment, so those images have the elements of ‘intellectual achievement’.”

Secondly, regarding the recognition of “originality”, the court held that, “The plaintiff used prompts to design visual elements such as characters and their presentation methods, and set parameters for the layout and composition of the picture, which reflected the plaintiff’s selection and arrangement. On the other hand, by inputting prompts and setting relevant parameters, the plaintiff obtained the first image, based on which the plaintiff continued to add prompts and modify parameters, with continuous adjustments, the plaintiff obtained the final images. This adjustment process also reflected the plaintiff’s aesthetic choices and personal judgments, …therefore, those images are not ‘mechanical intellectual achievements’. In the absence of counterevidence, it can be concluded that those images were independently completed by the plaintiff, reflecting the plaintiff’s personalized expression.

This case is the first copyright infringement case involving AI generated images in China. Prior to this, the Nanshan court in Shenzhen had ruled on a copyright infringement case involving AI generated articles ((2019) Yue 0305 Min Chu 14010). In this case, the court also determined that the article automatically generated by AI shall be deemed as “works”, it held that, “From the perspective of the generation process of the article, the expression form of the article is determined by the personalized arrangement and selection of relevant personnel of the plaintiff’s creative team, and its expression form is not unique and has a certain degree of originality.”.

However, there is great controversy in the legal and judicial practice regarding whether the content (images, articles, etc.) generated by AI constitutes works. The core issue lies in whether the content generated by AI has originality.

The author believes that based on the technical characteristics of AI, the originality of content generated by AI should not be recognized.

In judicial practice, the determination of “originality” usually takes two factors into account. They are, whether it was independently created by the author, and whether the arrangement of expression reflects the author’s choice and judgment.

As for the first factor, if users do not use AI, they cannot independently create corresponding “works”, so the users could not complete the “works” independently. Because the use of AI is different from the use of other technological means. For example, the tools from drawings from hand to rulers, and now we use computer, those tools are the technological means, which are used to express the author’s “thoughts”. The rejection of the copyright registration application of the “Space Opera House” in the United States reflects this viewpoint. Jason Allen used the AI Mid Journey system to create the “Space Opera”, which won an art competition award at the 2022 Colorado State Exposition. However, when Jason applied for copyright registration, the US Copyright Office deemed that the AI system had a substantial impact on the formation of the final works and rejected the application.

As for the second factor, the answer should also be negative.

Firstly, a prerequisite for AI to generate content is that AI has been “fed” with sufficient data/materials. In other words, AI is based on instructions issued by human beings to select, arrange, and combine existing data/materials, which is similar to assembly works in terms of its expression method. However, the content generated by AI clearly does not meet the definition and requirements of assembly works.

Secondly, consider whether the content generated by AI reaches a high level of intellectual creativity. As aforementioned logic, AI generated content is the arrangement and combination of existing data\materials, excluding non-works such as simple lines, factual messages, chart formulas, laws and regulations, or items that are not applicable to the “Copyright Law”, and the remaining parts are existing works. Therefore, for the arrangement and combination of existing works, even if the forehead is sweating, it usually cannot reach the level of “intellectual creation”.

Finally, in reverse deduction, if it is determined that the AI generated content is a work protected by the “Copyright Law”, then the “work” may be suspected of infringing on the copyright of the corresponding work of the original data\material. From the perspective of judicial outcome orientation, it may fall into an unfavorable cycle.

In summary, in view of the current legal provisions and the current stage of AI technology, we believe that it is not appropriate to recognize AI generated content as works which could be protected by the “Copyright Law”.