Authored By: TEO JANE
Multimedia University
The Rise of Machines, The Fall of the Author?
Artificial Intelligence (AI), a term first coined by McCarthy in 1955 as “the science and engineering of making intelligent machines” — has cast doubt on copyrightability of literary or artistic works generated by AI systems. The Copyright Act 1987 (hereinafter referred to as ‘CA 1987’), copyright protection is afforded to “original works of authorship.”, a standard traditionally interpreted as requiring human authorship for a valid claim. Infringement occurs when the newly created outputs resemble the existing copyrighted work, potentially involving the reproduction of the entire work or even substantial part thereof. Such issue exacerbated by the recent proliferation of generative AI tools such as ChatGPT, Gemini and Claude, capable of authoring books, composing music, creating artwork. This technological advancement necessitates a re-evaluation of authorship itself. Bad news is, Malaysia’s copyright regime remains grounded in human-centric authorship, but this paradigm appears unstable in the face of AI.
Decoding the Human-Centric Law — Malaysian Copyright Act 1987
The case of Honda Giken Kogyo Kabushiki Kaisha v Allied Pacific Motor (M) Sdn Bhd,[1] outlines four criteria for copyrightable works: (1) it falls within one of the categories of protected works; (2) it is original where sufficient skill and effort have been expended in the creation; (3) it fixed in tangible form; and (4) it fulfils the qualifications for copyright.
The concept of originality has been a longstanding one in the field of copyright. From a legal standpoint, the definition of originality in copyright differs greatly from the concept of novelty known by the general public. With reference to Article 2(1) of the Berne Convention,[2] it is clear that the standards, tests, or elements of determining originality will be a matter of national legislature and courts. To be qualified as protected work in Malaysia, sufficient effort ought to be expended for a creation to be original in character by virtue of section 7(3)(a) of CA 1987.[3]
In the case of Megnaway Enterprise Sdn Bhd v Soon Lian Hock, ‘sufficient effort’ refers to the degree of “effort, skill or labour” expended in the creation of the work.[4] Peterson J sets out the requirements in the case of University of London Press Ltd v University Tutorial Press Ltd that the word ‘original’ does not demand original or inventive thought but instead, the work should not be copied and must be originated from the author.[5] This draws a distinction line between the concept of originality in copyright law and the notion of novelty in the context of patent. Abdul Malik J further clarified the word through Kiwi Brands (M) Sdn Bhd v Multiview Enterprises Sdn Bhd, that ‘original’ does not mean the work must be the expression of original or inventive thought since idea is not protected under the prevailing copyright law.[6] For example, in the case of Nordic Water Products Aktibolag & Anor v Pumpen Environmental Sdn Bhd,[7] the plaintiffs claimed infringement of copyright for their water treatment equipment. It had to be determined whether their design drawings were original. The court found that the drawings took a period of 5 to 6 years to complete showing that their creation required a substantial amount of skill, labour, and effort.
In the general rule, a work can only be protected by copyright in Malaysia if the author has exercised sufficient intellectual effort independently in connection to Lockean’s theory. To generate outputs, AI machines will rely on a dataset of preexisting works, where some of them will undoubtedly cover copyrighted material. Section 8(1)(a) of CA 1987,[8] extends protection to derivative works such as adaptations and transformations of original materials. However, given that the outputs are typically generated by AI tools in mere seconds or minutes, with minimal “effort, skill or labour” from the user, by no means such outputs could reach the threshold needed for copyright protection under the Act.
AI as a Tool vs AI as a Creator: The Dividing Line
Works subject to copyright laws are now frequently created by autonomous algorithms, where the systems derive the data from the corpora which comprise the copyrighted work, to analyse the framework inherent to the human-generated data and thus synthesised similar data in the form of text, image, and audio. Section 7(1) of CA 1987[9] provided the works eligible for copyright which covers literary, musical, and artistic works. These works can be produced without human intervention with the tremendous advances in AI.
For literary works, the AI writes content through the software process that automatically creates a written narrative from data, which is called natural language generation (NLG). It determines the format content on the demand of the users and each type of content portrays a unique writing style and structure. The OpenAI’s GPT-2 model was trained through millions of documents for the text which was derived from various sources on the Internet and achieved state-of-the-art results by NLG.[10] The tools that have been widely used to generate textual context include Quill from Narrative Science, Wordsmith from Automated Insights, and Google’s Text-to-Speech.[11]
Music is another creative realm infiltrated by AI. The OpenAI MuseNet marked similar algorithms with GPT-2, where it integrated thousands of MIDI audio files and gave birth to a new MIDI file that carried its features without copying the published audio. In 2018, the release of ‘Hello World’ marked the first mainstream music album composed with AI, by the artist SKYGGE with the help of the Flow Machines project. In 2019, Warner Music signed a record deal with Endel to produce 20 albums to be put on streaming services and share an equal part of royalty with Endel. It provided a platform to generate personalised soundscapes based on surroundings. Those songs are owned by programmers of Endel.[12]
Under section 3 of CA 1987, artistic work includes graphic work and photography, irrespective of artistic quality.[13] Not only musical works but the resulting pieces of fine art and other creative products can be subjected to commercial use by the programmers or the users of the AI tools. DeepDream is an example of a computer program capable of analysing an image to produce a new dream-like visual art by enhancing the features of the original photograph.[14] Although initiated by human input, it tends to produce visual art that falls outside the realm of commonly produced human artwork. It reflects that the majority of AI visual art emphasises the role of prior images and the need for humans to guide the creative process.
Dilemmas in AI Authorship
Since it has been established beforehand that AI does not qualify as a natural person, here comes the grey zone: ‘Who is the de jure author over the content generated by AI if one heavily prompt, revise and re-edit it?’. The CA 1987 defines “author” in various contexts, such as a writer for literary works, a composer for musical works, and an artist for artistic works.[15] Yet, CA 1987 lacks explicit guidance for non-human authorship, placing the AI-generated works into a legal vacuum. Consequently, an analysis of potential authors revolves around two main possibilities: considering AI as the author, or designating the user, programmer, or AI company as the author.
Firstly, AI does not fall within the definition of the author in CA 1987 and lacks high degrees of autonomy, hence it needs a human author to be copyrighted. In Malaysia, the definition of “author” only applies to human authors under CA 1987.[16] Similarly, the US Copyright Office only acknowledges copyright for works created by humans. However, the authorship of AI-generated work is not guaranteed to the human “author” who may not exert meaningful control but merely provide only ideas. According to a legal scholar, a human user who inserts a text prompt into an AI program has contributed nothing more than a concept to the ultimate work; hence, the output works lack a human author and cannot be copyrighted. In 2023, ‘Zarya of the Dawn’ marked the latest development of copyrightability in the burgeoning field of AI-assisted creative work. It was a comic book created by Kris Kashtanova with the assistance of Midjourney, an AI-based image generator. In essence, Kris authored the storyline and employed Midjourney to produce visual elements through specific text prompts through multiple iterations. The completed work was the combination of Kris’s narrative with AI-generated illustrations. The US Copyright Office eventually granted protection for text and arrangement of images but denied protection for individual images within the graphic novel in absence of human authorship.
Furthermore, generative AI is not autonomous and self-aware on a human-like level to be considered an author. However, such a high standard is unlikely to be met at this point. This is supported by Andrew Wu, who suggests that awarding authorship to AI’s works may only be allowed under certain conditions, including producing unanticipated works, avoiding human involvement, and having the ability to determine when to create future works. Hence, although AI-generated works may meet the requirement for “free and creative choices,” a machine must possess a “human-like personality” to be considered an author in the EU.[17] Next, the human author is crucial for a valid copyright claim. In the case of Thaler v. Perlmutter, the court ruled that a work created using AI without a human author is ineligible for copyright protection.[18]
After conveying that AI cannot be the author, another question arises here: ‘Can the user, programmer, or AI company be deemed the author of AI-generated work?’ Unfortunately, the answer remains negative. In fact, neither the programmer nor the AI company are truly inserting creative inputs to deem them as authors. Their roles lack substantive involvement in the incentive process for AI-generated work, contributing minimally to the intellectual conception of the final output. Consequently, they do not meet the criteria for authorship.
At the heart of the AI authorship debate lies a legal paradox: AI’s work lacks legal personality, and thus ineligible for copyright, and nobody shall assert copyright towards it. Liability compounds the problem. If an AI system inadvertently or deliberately outputs infringing content — say, a piece of art mimicking the style of a famous painter — who is responsible? The days after OpenAI rolled out the latest feature on ChatGPT to date, it stunned users with ability to generate vast variety of images styles reminiscent to classic claymation. Unsurprisingly, it reignited a long-standing debate over copyright concerns over the viral trend of mimicking Japanese animation company Studio Ghibli through user’s prompts. AI tools are no longer stranger to art emulation and is considered as a form of flattery. Ironically, the co-founder Hayao Miyazaki described AI-generated art as an “insult to life itself”. The copyright laws however do not protect a visual style, and creation of fan art simply bears the risk of running afoul of such law but leaving nobody accountable for plagiarism, unless it has gone overboard for commercial utilisation and ran afoul of unfair competition. What’s clear is that these derivative and generative content are repeatedly stretching the limitations of copyright laws.[19]
Towards a harmonised legal framework on copyright
In conclusion, the rapid advancement of generative AI technologies would continually prompt intricate issues in future, indicating a need for a robust legal framework to tackle the public outcry. The absence of legal personality leaves the party concerned in limbo over their roles and responsibilities since the current regime is rooted in human authorship. Moving forward, copyright laws should act as a sword and shield for one to assert and defend his rights over contents that are AI-generated.
Bibliography
Primary Sources
Cases
Honda Giken Kogyo Kabushiki Kaisha v Allied Pacific Motor (M) Sdn Bhd [2005] 3 Malayan Law Journal 30 (HC)
Kiwi Brands (M) Sdn Bhd v Multiview Enterprises Sdn Bhd [1998] 6 Malayan Law Journal 38 (HC)
Megnaway Enterprise Sdn Bhd v Soon Lian Hock [2009] 3 Malayan Law Journal 525 (HC)
Nordic Water Products Aktibolag & Anor v Pumpen Environmental Sdn Bhd [2010] 7 All Malaysia Reports 523 (HC)
Thaler v. Perlmutter 2023 United States Reports (D.D.C. March 18, 2025)
University of London Press Ltd v University Tutorial Press Ltd [1916] 2 Chancery Division 601 (Ch D)
Statutes and statutory instruments
Berne Convention, art 2(1)
Copyright Act 1987
Secondary sources
Journal articles
Makhmutov M, Varouqa S and Brow JA, ‘Survey on Copyright Laws about Music Generated by Artificial Intelligence’ [2020] 2020 IEEE Symposium Series on Computational Intelligence (SSCI)
Petar Hristov Manolakev, ‘Works Generated by AI – How Artificial Intelligence Challenges Our Perceptions of Authorship’ (Tilburg University, July 2017)
Online journals
Cullen O’Keefe and others, ‘Before the United States Patent and Trademark Office Department of Commerce Comment Regarding Request for Comments on Intellectual Property Protection for Artificial Intelligence Innovation’ (2019) <https://www.uspto.gov/sites/default/files/documents/OpenAI_RFC-84-FR-58141.pdf> accessed 10 April 2025
Victor M. Palace, ‘What If Artificial Intelligence Wrote This? Artificial Intelligence and Copyright Law’ (2019) 71 Florida Law Review 217 <https://scholarship.law.ufl.edu/cgi/viewcontent.cgi?article=1439&context=flr> accessed 10 April 2015
Newspaper articles
Rowlands C, ‘Is ChatGPT’s Studio Ghibli Craze a Copyright Timebomb? Here’s the Verdict from Expert Lawyers’ (TechRadar3 April 2025) <https://www.techradar.com/computing/artificial-intelligence/is-chatgpts-studio-ghibli-craze-a-copyright-timebomb-heres-the-verdict-from-expert-lawyers> accessed 12 April 2025
[1] Honda Giken Kogyo Kabushiki Kaisha v Allied Pacific Motor (M) Sdn Bhd [2005] 3 Malayan Law Journal 30 (HC).
[2] Berne Convention, art 2(1).
[3] Copyright Act, s 7(3)(a).
[4] Megnaway Enterprise Sdn Bhd v Soon Lian Hock [2009] 3 Malayan Law Journal 525 (HC).
[5] University of London Press Ltd v University Tutorial Press Ltd [1916] 2 Chancery Division 601 (Ch D).
[6] Kiwi Brands (M) Sdn Bhd v Multiview Enterprises Sdn Bhd [1998] 6 Malayan Law Journal 38 (HC).
[7] Nordic Water Products Aktibolag & Anor v Pumpen Environmental Sdn Bhd [2010] 7 All Malaysia Reports 523 (HC).
[8] Copyright Act 1987, s 8(1)(a).
[9] Copyright Act 1987, s 7(1).
[10] In AI, “training” refers to the process by which an AI model learns patterns from the data input by its programmers.
[11] Cullen O’Keefe and others, ‘Before the United States Patent and Trademark Office Department of Commerce Comment Regarding Request for Comments on Intellectual Property Protection for Artificial Intelligence Innovation’ (2019) <https://www.uspto.gov/sites/default/files/documents/OpenAI_RFC-84-FR-58141.pdf> accessed 10 April 2025.
[12] Munir Makhmutov, Selina Varouqa and Joseph Alexander Brow, ‘Survey on Copyright Laws about Music Generated by Artificial Intelligence’ [2020] 2020 IEEE Symposium Series on Computational Intelligence (SSCI).
[13] Copyright Act 1987, s 3.
[14] Victor M. Palace, ‘What If Artificial Intelligence Wrote This? Artificial Intelligence and Copyright Law ’ (2019) 71 Florida Law Review 217 <https://scholarship.law.ufl.edu/cgi/viewcontent.cgi?article=1439&context=flr> accessed 10 April 2015.
[15] Copyright Act 1987, s 3.
[16] Copyright Act 1987, s 3.
[17] Petar Hristov Manolakev, ‘Works Generated by AI – How Artificial Intelligence Challenges Our Perceptions of Authorship’ (Tilburg University, July 2017).
[18] Thaler v. Perlmutter 2023 United States Reports (D.D.C. March 18, 2025).
[19] Christian Rowlands, ‘Is ChatGPT’s Studio Ghibli Craze a Copyright Timebomb? Here’s the Verdict from Expert Lawyers’ (TechRadar3 April 2025) <https://www.techradar.com/computing/artificial-intelligence/is-chatgpts-studio-ghibli-craze-a-copyright-timebomb-heres-the-verdict-from-expert-lawyers> accessed 12 April 2025.