Legal implications – Beijing Internet Court grants copyright protection to AI-generated artwork

I.   Introduction

Given the emergence of generative artificial intelligence (AI) as a staple in the toolkit of a modern-day creator, legal systems worldwide have found themselves grappling with the issue of whether copyright subsists in AI-generated works at an ever-increasingly frenetic pace. These decisions, which invariably make reference to each other as each legal system seeks to eke out its own principled way forward, will undoubtedly have a profound impact on creators’ approach to adoption of generative AI as an assistive tool going forward. In this evolving landscape, where AI-generated works exhibit increasing degrees of quality, sophistication, and are typically bespoke nature, the crucial question arises: do these outputs transcend the originality threshold, thus meriting copyright protection under existing legal frameworks? While many jurisdictions, have either determined that an AI-generated artwork is not subject to copyright, or have held that only works authored by a natural person (i.e. a human being) are deserving of copyright protection, recent developments in China appear to have taken a contrarian step in the opposite direction.

In the recent decision of Li v Liu,[1] the Beijing Internet Court was asked, once again, to recognise the copyrightability of an AI-generated artwork.

This article will summarise the decision in Li v Liu and explore how and why the Beijing Internet Court reached a different conclusion from the USCO, given a similar set of facts. Finally, we will briefly explore the implications of Li v Liu and Zarya of the Dawn[2] on copyright in AI-generated content going forward.

II.  A new era of art and artistry: The background of Li v Liu

On 24 February 2023, the plaintiff, Mr Li Yunkai, generated the picture involved by inputting prompts in Stable Diffusion, an open source software, and then published the picture on Xiaohongshu, a social media platform, under the title “Spring Breeze Brings Tenderness”. The plaintiff later discovered that the defendant, Ms Liu Yuanchun, a blogger on the Baijiahao platform, had used the picture in an article titled “Love in March, in the Peach Blossoms”, published on 2 March 2023 without his permission.

Figure 1: Picture generated by Li, titled “Spring Breeze Brings Tenderness”

The Beijing Internet Court began by setting out the elements that must be satisfied for a work to attract copyright protection, one of which is that the work must be “original”.  As the Court opined: [3]

Of course, not all intellectual achievements are works; only those with “originality” are. Generally speaking, “originality” requires that the work be completed independently by the author and reflect the author’s personalized expression.

The Court went on to explain:[4]

Generally speaking, when people use the Stable Diffusion model to generate pictures, the more different their needs are and the more specific the description of picture elements, layout, and composition is, the more personalized the picture will become. In this case, there are identifiable differences between the picture involved and the prior works. In terms of the generation process of the picture involved, the plaintiff did not draw the lines himself, or instruct the Stable Diffusion model everything on how to draw the lines and do the colors; the lines and colors that constitute the picture involved are basically done by the Stable Diffusion model, which is very different from the conventional way of people using brushes or software to draw pictures. However, the plaintiff used prompt words to work on the picture elements such as the character and how to present it, and set parameters to work on the picture layout and composition, which reflects the plaintiff’s choice and arrangement. The plaintiff input prompt words and set parameters and got the first picture; then he added some prompt words, modified the parameters, and finally got the picture involved. Such adjustment and modification also reflect the plaintiff’s aesthetic choice and personal judgment. [emphasis added in bold]

Accordingly, the Court held that the plaintiff’s AI-generated work was original.

A further notable quirk of Chinese copyright law is the requirement for “intellectual achievement”, which, according to the Court in Li v Liu, refers to the “results of intellectual activities”. In other words, “the work should reflect the intellectual input of a natural person”. [5] The Court noted that the plaintiff did indeed incur a degree intellectual investment, including “designing the presentation of the character, selecting prompt words, arranging the order of prompt words, setting parameters, and selecting the picture that he wanted”. [6] The picture therefore reflected the plaintiff’s intellectual investment, such that the element of “intellectual achievement” was met.

Interestingly, on the topic of “mechanical intellectual achievements”, the Court held that “’[m]echanical intellectual achievements’ are excluded [from copyright]. For example, if a work is completed based on a certain order, formula, or structure, different people will get the same result; as the expression is singular, the work does not have originality”.[7] However, the Court reasoned that Spring Breeze Brings Tenderness was not a “mechanical intellectual achievement”, because “different people can generate different pictures by entering different prompt words and setting different parameters”. [8] This may be specific to the facts of this case, as one would expect that different people entering the same parameters and prompts and using the same seed (i.e. the randomly generated number that determines how the AI generates the initial white noise image which will be subsequently refined in accordance with the user’s prompts) in Stable Diffusion would indeed generate the exact same picture that the plaintiff had arrived at. For example, an online resource explains that “[t]he most important thing about seed is that generations with the same parameters, prompt, and seed will produce precisely the same images”.[9] An AI-generated image that is “completed based on a certain order, formula, or structure”, such that “different people will get the same result”, would, according to the Court’s definition, make the work unoriginal under Chinese law. This also raises the question as to what degree of complexity of prompts would be required to take an AI-generated image out of the realm of a mechanical intellectual achievement, given that the Beijing Internet Court’s approach appears to be largely dependent on whether the prompts used result in a sufficiently differentiated work.

In particular, the Court, in acknowledging the plaintiff as the copyright owner of the AI-generated work, noted that “generative AI technology has changed the way people create. Just like many other technological advances in history, the process of technological development is the process of outsourcing human work to machines… [t]he 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”.[10]  In terms of copyright ownership, the Court likened a user’s use of generative AI to a consignment to a painter to paint for the user, with the exception that “the [painter] 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, the not AI model”.[11]

Finally, the Court concluded that recognising the copyrightability of AI-generated works would help to encourage creation using the latest technology: [12]

The core purpose of the copyright system is to encourage creation. And creation and AI technology can only prosper by properly applying the copyright system and using the legal means to encourage more people to use the latest tools to create. Under such context, as long as the AI-generated images can reflect people’s original intellectual investment, they should be recognized as works and protected by the Copyright Law.

III.  The Road Not Taken: Comparison with Zarya of the Dawn

In the case of Zarya of the Dawn in the United States, Kristina Kashtonova had used Midjourney (a different AI platform) to produce images based on text prompts to create illustrations for her graphic novel, “Zarya of the Dawn”.  Like the plaintiff in Li v Liu, Ms Kashtonova arrived at the final pictures through a process of trial-and-error, as she provided “hundreds or thousands of descriptive prompts” to Midjourney until the “hundreds of iterations [created] as perfect a rendition of her vision as possible”.[13]

The USCO took the view that the copyright registration was limited only to the text and selection and arrangement of the written and visual elements, but not the individual images generated by Midjourney.  Per the USCO, it was trite (in US law) that only human authors can claim copyright protection for their works.  However, Ms Kashtanova was not considered the author of the Midjourney-generated images:[14]

Rather than a tool that Ms. Kashtanova controlled and guided to reach her desired image, Midjourney generates images in an unpredictable way. Accordingly, Midjourney users are not the “authors” for copyright purposes of the images the technology generates…. Because of the significant distance between what a user may direct Midjourney to create and the visual material Midjourney actually produces, Midjourney users lack sufficient control over generated images to be treated as the “master mind” behind them. [emphasis added in bold]

Indeed, the USCO distinguishes between the use of generative AI and the use of other tools in the creative process, on the basis that unlike other kinds of creative tools, users using generative AI have insufficient control over and are unable to predict or guarantee the outcome of generative AI’s output.  This contrasts with the view of the Beijing Internet Court, which opined that the prompts given by Li “reflect [his] aesthetic choice and personal judgment”.

Further, the USCO similarly likened the use of generative AI to a user commissioning a “visual artist” to produce an image, with the prompts functioning “closer to suggestions than orders”. The USCO thus opined that the author in this case would not be the commissioner but “the visual artist who received those instructions and determined how best to express them”.[15] Making a similar analogy, the Beijing Internet Court, however, took the view that there is “no question” that the generative AI user is the creator, because generative AI “has no free will and is not a legal subject”. [16]

Finally, whilst Ms Kashtanova had expended significant time and effort working with Midjourney, the USCO reiterated that effort does not make her the “author” of the images and cannot be the basis for copyright protection in otherwise unprotectable material.  Nonetheless, the Court left the door open to allowing copyright protection in a future case based on a different set of facts, opining that it is “possible that other AI offerings that can generate expressive material operate differently than Midjourney does”.[17] 

IV.  Concluding Comments

Overall, both the USCO and the Beijing Internet Court appear to have raised compelling arguments in coming to their respective decisions. The question of whether AI-generated works satisfy the originality requirement is a matter of ongoing discussion. Indeed, in many legal systems such as Singapore, the current understanding of originality in copyright law emphasises human intellectual creation (such as for compilations)[18] and human authorship,[19] whilst also prohibiting slavish copying.[20] This raises the question of whether works qualify as "original" when human involvement is limited to providing prompts or instructions, essentially acting as an intermediary for the AI's creative output.

The decision in Li v Liu therefore represents a noteworthy breakaway from the main corpus of law relating to the subsistence of copyright in AI-generated works.  There could be a myriad of reasons for this unique outcome, although one might venture a possible interpretation that under Chinese law, the requirement for human authorship does not appear to feature in the inquiry as to “originality”. According to the Beijing Internet Court, “originality” and “intellectual achievement” are separate requirements that must be met before copyright can subsist in a work under Chinese law.[21] To this end, “intellectual achievement” refers to “the results of intellectual activities, so the work should reflect the intellectual input of a natural person” (emphasis added).[22] On the other hand, the Chinese approach to “originality” merely requires that the resulting work is “completed independently by the author and reflect the author’s personalized expression”.[23] In other words, the Chinese test for “originality” and “intellectual achievement”, when read together, do not expressly require the author to exercise human creativity, but instead appear to emphasise whether a work is completed independently and involves some form of intellectual activity (which appears to include the intellectual input of a human exercising the functions of an AI, as in this case), such that the end result is a work which reflects the author’s personalised expression.

The element of “intellectual achievement” does not appear to be a pre-requisite for copyright to subsist in the other jurisdictions that have grappled with this question. Instead, the requirement for originality is phrased as one of “intellectual creation", which jurisdictions have interpreted as requiring the exercise of human creativity. For example, as stated by the US Copyright Office in its Copyright Registration Guidance on Works Containing Material Generated by Artificial Intelligence, “it is well-established that copyright can protect only material that is the product of human creativity” – as such, the Office “will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author”.[24] Given the emphasis on human creativity in the test, it is understandable that most major jurisdictions have thus far declined to recognise the copyrightability of AI-generated works.

There is little doubt that courts will continue to struggle with finding the right balance relating to the protection of generative AI works moving forward. Given the lack of judicial consensus and the difficulty of fitting generative AI into the traditional copyright regime, the decision is likely to come down to policy rather than law.  Alternatively, sui generis legislation may have to be implemented.

For now, the Beijing Internet Court appears to have adopted Robert Frost’s wisdom:

Two roads diverged in a wood, and I—
I took the one less traveled by,
And that has made all the difference.

It may indeed make all the difference in the field of AI-generated copyright in days to come.

AUTHOR INFORMATION

Dr Stanley Lai, SC is a Partner at Allen & Gledhill, where he is also the Head of the Intellectual Property Practice and Co-Head of the Cybersecurity & Data Protection Practice.
Email: stanley.lai@agasia.law

David Lim is a Senior Associate in the Intellectual Property Practice of Allen & Gledhill.
Email: david.lim@agasia.law

Linda Shi is an Associate in the Intellectual Property Practice of Allen & Gledhill
Email: linda.shi@agasia.law

Justin Tay is an Associate in the Intellectual Property Practice of Allen & Gledhill.
Email: justin.tay@agasia.law

REFERENCES

[1]            Beijing Internet Court Civil Judgment (2023) Jing 0491 Min Chu No. 11279.

[2]            Zarya of the Dawn (Registration # Vau001480196) (link here) (accessed 6 February 2024).

[3]           Supra note 1, p. 11.

[4]            Supra note 1, p. 12.

[5]            Supra note 1, p. 10-11.

[6]            Supra note 1, p. 11.

[7]            Supra note 1, p. 11-12.

[8]            Supra note 1, p. 12.

[9]            https://getimg.ai/guides/guide-to-seed-parameter-in-stable-diffusion (accessed 6 February 2024).

[10]           Supra note 1, p. 12-13.

[11]           Supra note 1, p. 13.

[12]           Supra note 1, p. 13.

[13]           Supra note 2, p. 8.

[14]           Supra note 2, p. 9.

[15]           Supra note 2, p. 10.

[16]           Supra note 1, p. 13.

[17]           Supra note 1, p. 13.

[18]        See Global Yellow Pages Ltd v Promedia Directories Pte Ltd [2017] 2 SLR 185.

[19]        See Asia Pacific Publishing Pte Ltd v Pioneers & Leaders (Publishers) Pte Ltd
[2011] 4 SLR 381.

[20]        See Auvi Pte Ltd v Seah Siew Tee [1991] 2 SLR(R) 786.

[21]        Supra note 1, p. 10-11.

[22]        Supra note 1, p. 10-11.

[23]        Supra note 1, p. 11.

[24]        https://copyright.gov/ai/ai_policy_guidance.pdf, p. 2-3 (accessed 20 February 2024).