Diverging International Approaches to the Copyrightability and Authorship of AI-created Works

When can a work created using artificial intelligence ("AI") be copyrighted? And, if an AI-enabled work can be copyrighted, who is the "author"? The Beijing, China Internet Court took a different approach than the U.S. Copyright Office, granting copyright protection to images created using AI – while stopping short of treating the AI itself as an author.[1]

The U.S. Approach: Zarya of the Dawn

First, some background: In the United States, the Copyright Office last year famously denied copyright protection to pictures in a comic book, Zarya of the Dawn, after it learned that the pictures had been generated using Midjourney, an image-generating AI.[2] The artist, Kristina Kashtanova, used an iterative process to create the images. She gave Midjourney an initial prompt, which produced several images; she selected the one closest to what she wanted; she provided more prompts to cause Midjourney to generate more images (based on the one she selected), and then she repeated this process until she had the images she wanted for the comic book.[3]

After the registration was approved, the Copyright Office became aware that Kashtanova had relied on Midjourney to create some or all of the images contained in Zarya of the Dawn. The Copyright Office preliminarily concluded that the information in Kashtanova's application was "incorrect or, at a minimum, substantively incomplete"[4] and initiated cancellation proceedings because, by Kashtanova's own admission, she was "not the sole author of the entire work and, at a minimum, the claim should have been limited to exclude non-human authorship."[5] The Copyright Office then cancelled registration for the images, relying on well-established precedent to conclude that under United States law, to be subject to copyright, a work must have a human author – which Midjourney, obviously, was not.[6]

But the Copyright Office went further than that. It found that even though Midjourney could not be the author of the pictures for copyright purposes, neither could Kashtanova. Its reasoning was that while Kashtanova had in some sense "told" Midjourney what to do, the actual creation of the pictures was accomplished by Midjourney, not by Kashtanova, which meant that she was not the author of the pictures either. Notwithstanding the creativity and artistic judgment that went into Kashtanova's repeated selection of prompts and choosing the image (out of several that Midjourney generated each round) that best fit her artistic vision, the actual creation of the images was done by an AI program, not by a human. The Copyright Office determined that Kashtanova had begun the image generation process with a "field of visual 'noise'" provided my Midjourney; had "influence[d]" the generated image through her prompts, but had not exercised sufficient control over the image "to be treated as the 'mastermind' behind them."[7] The result is that – at least in the view of the Copyright Office – images created by an AI using Kashtanova's iterative process can have no "author," and thus no copyright protection.[8]

The Chinese Approach: Li v. Liu

This brings us to Li v. Liu. In this case, Mr. Li used Stable Diffusion, another image-generating AI model, in a manner similar to the way Ms. Kashtanova used Midjourney in the Zarya of the Dawn case: he started with some initial prompts, looked at the image Stable Diffusion generated in response, and continued to modify and refine the prompts until he got the picture he wanted;[9] he then posted the final image online. The case arose when Ms. Liu obtained a copy of the picture and used it without any attribution or compensation to Mr. Li. Mr. Li sued, seeking an apology, removal of the picture from Ms. Liu's online site, and monetary compensation.

If this case had arisen under United States law, the result would likely have been the same as in Zarya of the Dawn. And indeed, the court in Li v. Liu affirmed that Stable Diffusion, the AI, was not a legal person and so could not be the author of a copyrighted work. To this extent, United States and Chinese law appear to be in accord.

With respect to copyrightability, as Zarya of the Dawn shows, under United States law, without an author there can be no copyright. Chinese law, however, approaches the question from a slightly different angle, holding that for a work to be copyrightable, it must be "original" and "an intellectual achievement." The court had no trouble finding that Mr. Li's image was "original" – it did not exist before his efforts with Stable Diffusion. The court also had no trouble finding that the work represented "an intellectual achievement" by Mr. Li – the record showed that he repeatedly considered what parameters to input into Stable Diffusion in order to have the program generate the final image he wanted.

With respect to authorship, the court considered whether to deem the designer of the AI an author, but rejected that view because the designer "neither had the intention to create the picture involved, nor did he preset the content generated afterwards."[10] The designer had merely produced a "creation tool," not the image itself.

That left Mr. Li. The Court noted that Mr. Li "is the one who directly set up the AI model involved as needed and finally selected the picture involved. The picture involved is generated directly due to the plaintiff's intellectual investment and it reflects the plaintiff's personalized expression. Therefore, the plaintiff is the author of the picture involved and owns the copyright on it."[11]

In this situation, Stable Diffusion should be considered to be a sophisticated tool that human creators could use to embody their human artistic and intellectual ideas in tangible form. From the opinion:

The development of technologies and tools require less human investment, but the copyright system should remain in use in order to encourage the creation of works. Before the emergence of the AI model involved, people needed to spend time and energy learning how to paint, or to consign others to paint for them. In the second scenario, the painter will draw the lines and fill in the colors upon the client's request to complete a work of fine art. And the person who draws is normally considered a creator. This is similar to the use of AI models to generate pictures, but there is one major difference here: the creator has his own will and he will use some judgment when painting for the client. Currently, the generative AI model has no free will and is not a legal subject. Therefore, when people use an AI model to generate pictures, there is no question about who is the creator. In essence, it is a process of man using tools to create, that is, it is man who does intellectual investment throughout the creation process, [not the] AI model.[12]

Having found that the image was copyrightable, and that Mr. Li was the author, the court then considered whether Ms. Liu's use of the image, without attribution or compensation, was an infringement, and readily concluded that it was. Ms. Liu had violated Mr. Li's rights under Chinese copyright law – specifically the rights of authorship and the right of "dissemination on the information network" – and the court therefore required Ms. Liu to issue an apology and pay 500 yuan to Mr. Li.

Closing Comments

From one perspective, the court's treatment of the AI-generated work as copyrightable and of the AI user, Mr. Li, as the author, makes sense: it is not unreasonable to recognize that AI models are a newly developed kind of tool that allow people to express their own originality and creativity. Indeed, this view is consistent with other nations' treatment of similar issues. For example, in British Commonwealth countries, the law specifically states that for a "work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken"[13] – in this case, Mr. Li. And, while there is no decision from EU courts of which we are aware on this specific topic, scholarly commentary on EU law suggests that the result under EU law would be in accord with the decision in Li v. Liu and Commonwealth-country law.[14]

So, neither Chinese law, as illustrated by Li v. Liu, nor United States law, as illustrated by Zarya of the Dawn, nor U.K. or other Commonwealth country laws, nor EU law, has yet crossed the Rubicon of treating an AI as an author of a work.

However, viewing the two cases together shows that they diverge in their approaches to whether and when the AI user should be treated as the author of the work. They involved similar processes, in each of which the user entered extensive prompts to shape the final image. The Copyright Office determined that Kashtanova had not exercised sufficient control to be the "master mind" of the work and thus could not be deemed the author (and the work, lacking a human author, could not be copyrighted). By contrast, the Beijing Internet Court found that plaintiff had made "intellectual investment" and the picture reflected his "personalized expression" sufficient to render him the author and copyright owner.

Unless the Copyright Office changes its approach (whether on its own or based on federal-court decisions), the United States will be a bit of an outlier in that a work generated by AI, even after extensive input and refinement by a human, may not be copyrightable at all if the human fails to exercise sufficient control to be the "master mind" of the final work. The more common view around the globe appears to be that even where the AI's contribution is quite substantial, the AI is seen as a tool facilitating the creation, by human authors like Kashtanova and Li, of new copyrightable works.