“Originality” Of Artificial Intelligence-Generated Art
“We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding” – this was the introduction by Google Brain Team to their new Artificial Intelligence (AI) art generators, Imagen, joining DALL-E, Midjourney, and Stability Diffusion in the new wave of AI art generators. However, the birth of AI-generated art (AI art) has generated controversies. One of the arguments raised in the war between AI corporations and artists is that – since AI art generators merely digest images from other artists/creators, it is unable to create something “original”. However, is this statement accurate under Malaysian copyright laws? Is AI art truly “unoriginal”?
Analysis
Before delving into the legal analysis of the “originality” of AI art, it is crucial to understand how AI art generators work. Christie’s, who recently auctioned off an AI art titled ‘Portrait of Edmond Belamy’ for US$ 432,500, described the painting as a creation of AI, “an algorithm by [an] algebraic formula”. To operate an AI art generator, a vast dataset of artworks is “fed” into the algorithm for it to “learn” the aesthetics and artistic components (input). The algorithm then attempts to emulate the input and generate a range of “new” images based on the prompts keyed into the program.
At face value, AI art would constitute a type of “artistic work” eligible for copyright protection under Section 7(1) of the Copyright Act 1987 (CA). However, to determine if AI art is original, two main considerations should be made: (a) who is the author of the AI Art? (b) has sufficient effort been expended to create a work that is original in character?
Section 3 of the CA defines the “author” of an artistic work as the “artist” of that work. In Dura-Mine Sdn Bhd v Elster Metering Ltd & Anor [2015] 3 MLJ 1, the Federal Court stated that for a work to be original, “the work must originate from the author.” The High Court in Creative Purpose Sdn Bhd & Anor v Integrated Trans Corp Sdn Bhd & Ors [1997] 2 MLJ 429 also indicated that Section 3 ought to be read together with Section 10(1) of the CA, in that “author” and “qualified person” share a common meaning. Thus, based on the present Malaysian copyright laws, we could deduce that the author of an “artistic work” would be an ‘artist’ who could either be - (a) an individual who is a Malaysian citizen or permanently resides here or (b) a body corporate established in Malaysia.
Evidently, if AI were human, it would qualify as the author as the CA protects the “originality” of the expression, not the novelty of the idea or portions thereof. As stated by the Federal Court in YKL Engineering Sdn Bhd v Sungei Kahang Palm Oil Sdn Bhd & Anor [2022] 8 CLJ 32, which also cited Dura-Mine Sdn Bhd in approval, the copyright law “does not require that a work should be unprecedented or new in comparison with already existing works”; in fact, “it was observed that a creator or author of an original work is entitled to copyright even though he may have infringed the copyright of any work which he has drawn or copied from.”
To illustrate, reference is made to the analogy in Dave Grossman Designs, Inc. v Bortin 347 F. Supp. 1150 (N.D. Ill. 1972):
"The law of copyright is clear that only specific expressions of an idea may be copyrighted, that other parties may copy that idea, but that other parties may not copy that specific expression of J 429 also indicated that Section 3 ought to be read together the idea or portions thereof. For example, Picasso may be entitled to a copyright on his portrait of three women painted in his Cubist motif. Any artist, however, may paint a picture of any subject in the Cubist motif, including a portrait of three women, and not violate Picasso's copyright so long as the second artist does not substantially copy Picasso's specific expression of his idea."
Applying that analogy to this scenario, whilst Van Gogh would be entitled to his Expressionism-styled painting of irises, any artist, including the ‘AI’ may paint a picture of any subject in the same Expressionism style, so long as no substantial mimicry is done to the said specific expression of Van Gogh. On this basis, the argument that AI art is “unoriginal” because it merely “ingests” images from other artists does not hold water, as even human authors can be inspired and emulate the styles and composition of other artworks to produce an entirely new and unique expression of work.
However – AI is merely an algorithm/machine; hence, who is to be deemed as the author in this scenario? Unfortunately, this question remains unanswered due to the dearth of discussions and case law in Malaysia and the silence of CA on this matter.
One possible approach to consider is to view AI as a “tool”, utilized by a human author via the input of prompts to create their “vision”. An analogy could perhaps be drawn from photographs, in which Section 3 of the CA provides that the author of the photographs means, “the person by whom the arrangement for the taking of the photographs was undertaken”, similar to UK’s Section 9(3) and Section 178 of the Copyright, Designs and Patents Act 1988, which provide that “authors” of “computer-generated works” which was defined as works, not without a “human author”, shall be taken as the “person by whom the arrangements necessary for the creation of the work are undertaken”. Interestingly, when cameras were first introduced, arguments were made that “a photograph is not … a production of an author” as it is the “creation” of the camera (see Burrow-Giles Lithographic Co. v Sarony 111 U.S. 53, 56 (1884)), which uncannily resembles the arguments against AI in that AI art the “creation” of the AI. Back then, dissents against photographs had been rebutted by the Court in Burrow-Giles which held that photographs are still subject to copyright “as they are representatives of original intellectual conceptions of the author”. Could the same perhaps be applied to the person inserting prompts into the AI art generator?
Perhaps we could also draw inspiration from Wedding Galore Sdn Bhd v Rasidah Ahmad [2016] 6 CLJ 621 where the High Court held that the edited photograph (taken by the photographer herself) was eligible for copyright protection and the photographer who edited those photos was the copyright owner as her usage of Adobe Photoshop to enhance the colour and background, remove imperfections of the photographs was not too “minimal”. Through this case, we can deduce that the photographer used two tools – a camera and Adobe Photoshop to create the original image and the final edited product. Arguably, AI art generators are also a “tool” to create new images and/or to manipulate/enhance existing images.
The main question would be – what is the degree of ‘human touch’ required in AI art for it to be deemed “original”? In Malaysia, one of the requirements for copyright protection to arise in an artistic work is when sufficient effort has been put in to make the work original (Section 7(3)(a) of the Copyright Act 1987). This is a low threshold. So long as it is a product of work arising from new/original skill and labour, the work would be copyrighted (see Megnaway Enterprise Sdn Bhd v Soon Lian Hock (sole proprietor of the firm Performance Audio & Car Accessories Enterprise) [2009] 3 MLJ 525)
However, we need to keep in mind that AI art can be generated through simple acts, like inputting a general prompt and leaving it entirely to the AI’s “creativity” to generate results. For example, via DALL-E, the AI art tool developed by OpenAI, users can type in even the simplest prompt, like “pencil sketch of a dog with sunglasses”, and DALL-E will generate four AI images to be chosen by the user.
On the flip side, AI art can also be produced using “very specific prompt(s), generates many images”, and the author further curates "those images and carries out further edits” to produce the final image, which would more likely make authorship over the AI art more justifiable.[1]
Conclusion
At the end of the day, this requirement is an “abstract” one. How the AI art generator is used and how much effort is to be invested by the human author to create the final work, is to be decided on a case-to-case basis since it would be almost impossible to fix a “formula” to gauge one’s effort, skill or labour. Till then, it seems like at face value, AI art can be considered a piece of “original” work, but whether it could be protected under copyright depends on the degree of “human” effort placed into the final product for there a “human author” to be recognised.
Undoubtedly, copyright issues on AI art would require in-depth analysis and studies, especially since AI is here to stay and will continuously evolve and assimilate into various industries – whether we like it or not. As such, studying and developing laws/regulations to address and keep up with the ever-evolving AI technology is important and is an area worth dipping our toes into.
[1] https://www.reuters.com/article/tech-art-copyrights-idUSL8N2ZL5OZ
22 May 2023