Two Cases from Internet Courts in China concerning Copyright Aspects of AI-Generated Content

There has been a significant amount of debate about whether AI-generated works should enjoy copyright protection or not. Generally, a work may enjoy copyright protection if it is

1. A literary, dramatic, musical, or artistic work;
2. If the work was an original expression in a fixed material form.

Even if the same argument is made for AI-generated content, the question remains as to whom will the copyright protection be attributed. The programmer? The AI itself or the user of the AI whose prompt inputs resulted in the work? The US Copyright Office has declared that an original work of authorship, created by a human being will be registered only. While case law argues that copyright law protects the results of intellectual labor stemming from the creativity of the mind and that work generated by a computer will not attract copyright because it was not made by a human (Acohs Pty Ltd v Ucorp Pty Ltd).

Attaching copyright to the programmer and the AI itself have both been regarded unlikely. The perception of AI as a “tool” and the user of the AI enjoying copyright over the resulting work generated by the AI, is a proposition that has theoretically been supported. In 2023, The Beijing Internet Court issue issued a ruling that might be used in future litigation concerning user copyright over AI-generated work.

The facts are as follows, the plaintiff, one Mr. Li, on 24^th February, 2023, posted an AI-generated image on Xiaohongshu, a renowned Chinese e-commerce and social media site. He titled the image “Spring Breeze Brings Tenderness”. The image was generated through an open-source AI software named “Stable Diffusion” through prompt input. He wanted to draw a close-up portrayal of a woman in photographic style under sunset lighting conditions. On Stable Diffusion, he used prompts such as “super realistic photo”, and “color photo” to outline art type; “Japanese idol” to determine the subject alongside their eye, skin tone, and hair color; described the style as “film texture” and “film simulation”; the environment as “golden time”, “exterior”, and “dynamic lighting”; and the pose of the subject as “looking at the camera”, and “cool pose”.

The defendant, one Ms. Liu, on 2^nd March 2023, posted Mr. Li’s photo on Baijiahao, a content creation platform owned by Baidu, without Mr. Li’s permission, removing the original watermark/insignia. Upon discovery, Mr. Li filed a lawsuit against her for copyright infringement, demanding a 5,000 yuan compensation and an apology.

Ms. Liu argued that it was unsure whether the AI-generated image originally posted by Mr. Li had any copyright attached to it, and that her post was an original poetry without any commercial purpose.

The Beijing Internet Court found in favor of Mr. Li, and held that Ms. Liu infringed his copyright associated with that image. It held that

• The image met the requisite “intellectual achievement” standard.
• It met the “originality” threshold.
• It is a work of art.
• The copyright to this image is enjoyed by the author and it should be recognized as a work that is protected by copyright.

The Court, referred to Article 3 of the People’s Republic of China Copyright Law that defined the term “works” as “intellectual achievements in areas such as literature, arts, and sciences, that have originality, which can be fixed in certain forms”. The Court held that Mr. Li had “made a certain amount of intellectual investment,” in the final result by setting the parameters that resulted in content or “work” finally generated thus meeting the requisite “intellectual achievement” standard. The Court concluded by stating that in the past, artists had to produce artwork based on individual skills and ideas, but with the advent of AI, investment in creative work has shortened temporally, but it does not mean intellectual investment in the creation process of work generated by AI should not receive the benefit of copyright protection. Encouraging creation is the recognized core purpose of the copyright mechanism.

After the verdict was delivered by the court on 27^th November, Ms. Liu apologized within 24 hours and paid the compensation to Mr. Li within a week of the verdict.

This verdict by the Beijing Internet Court may well be the first concerning a legal dispute regarding user copyright over AI-generated content, but it will not be the last one.

AI Companies and their Liability for Copyright Infringement

To determine whether or not there has been an infringement of a copyrighted work, several factors are taken into consideration to determine infringement

1. Unauthorized Copy of original work expressed in a material or fixed form;
2. Copying included the original expression of the work and a substantial amount of it;
3. Whether or not the copying is covered by fair dealing;
4. Whether any other factors may operate as an exception to the copying of the original work.

In the case of AI companies and their liability for copyright infringement, when the copying occurred, i.e., i) during training the AI, or ii) for the creation of a new work, and whether it took place on multiple occasions are additional factors that need to be considered. In early 2024, the Guangzhou Internet Court took these exact factors into consideration in a copyright infringement dispute, holding that the defendant (an AI company) had infringed the copyright of the plaintiff concerning the reproduction and adaptation of Ultraman in generative AI works.

Ultraman or the Ultra Series originally started as the Ultra Q in 1966 about fictional alien superheroes fighting other aliens known as kaiju (similar to Pacific Rim), is a sci-fi media franchise owned by Tsuburaya Production Company Ltd. Tsuburaya signed a “Certificate of Authorization” with the plaintiff company providing them (the plaintiff) an exclusive license to the copyright of the Ultraman image series including the right to defend its rights.

The defendant is a company that operates the Tab (pseudonym) website providing AI dialogue and AI-generated drawing facilities. The AI-generated painting facility on the Tab website is a member-exclusive function requiring users to replenish “computing power” depleted in the course of producing the AI-generated content. The plaintiff’s lawsuit alleged that the defendant had used the works to which the plaintiff had rights to train its AI model (training data) and made substantial similar pictures without any authorization while making financial gains in the process. Thereby, causing considerable damage to the legitimate rights and interests of the plaintiff.

The Court held that there was a copyright infringement on the part of the defendant and ordered that 10,000 Yuan be paid to the plaintiff by the defendant. The Court reasoned that

1. The work produced by the Tab website had substantial similarity with the original artistic image of the work (i.e., Ultraman) through the retention of several key features and the original expression of the work. Therefore, establishing substantial similarity, constituting duplication of the plaintiff’s work without permission and resulting in infringement of the plaintiff’s right to reproduction of the work.
2. The work generated by the Tab website retained the original expression of the work and formed new characteristics based on the original expression. This amounted to an adaptation of the original work without the permission of the defendant, thus, infringing the right of the plaintiff concerning the right to adapt the work.

The Court went further and outlined obligations for companies providing generative AI services. It stated that users of such service providers should be reminded not to infringe the copyright of others through service agreements or other means, that service providers should establish a reporting mechanism that allows copyright holders to protect their rights through reporting, use of marking scheme in a reasonable location or area on AI-generated content to make it distinguishable. Acknowledging that AI is “..a strategic technology leading the future..” and that it “..is considered to be the main position for the development of new quality productivity”, it is needed to not overburden the obligations of service providers and strike a balance between rights protection and industrial development.

As of February 24, 2024, there have been more than 10 copyright infringement disputes related to generative AI in the United States, including Sarah Silverman v. OpenAI, Inc., Paul Tremblay v. OpenAI, Inc., Concord Music Group, Inc. v. Anthropic PBC, The New York Times Co. v. Microsoft Corp., OpenAI, etc. It will be interesting to see what the courts in the United States decide concerning these disputes and whether we will see a divergence in interpretation and approach from courts across jurisdictions concerning copyright disputes over AI-generated content.

Ahmed Ragib Chowdhury
LLM Student, Peter A. Allard School of Law,
The University of British Columbia.