Generative AI & Copyright Law in Malaysia: Challenges and the Way Forward

by Celine Lim Tze Hui, Justin Tee Keng Ghee, Herman Goh Hong Mun and Yau Khai Zen

A. Introduction

The rise of generative artificial intelligence (‘AI’) has sparked both excitement and legal uncertainty, particularly in the realm of copyright. Tools like ChatGPT, Midjourney, and Stable Diffusion are enabling users to produce text, images, music, and videos at unprecedented speed, often blurring the lines between human creativity and machine-generated output. As AI becomes more capable of creating sophisticated, human-like works, core copyright questions around authorship, originality, and ownership are being tested worldwide.

Some jurisdictions have started confronting these challenges head-on, while others, including Malaysia, are still in the early stages of grappling with the legal implications. Although Malaysia has introduced governance initiatives such as the National Guidelines on AI Governance and Ethics (‘AIGE’), the current legal landscape remains largely untested, with no clear judicial or legislative position on AI-generated works.

This article examines how Malaysia’s copyright framework interacts with generative AI. It outlines the key statutory requirements for protection, explores emerging legal challenges, and highlights how different jurisdictions, including the United States and China, are addressing the issue. Finally, it considers how Malaysia might approach reform, ensuring that copyright law continues to protect human creativity without stifling technological advancement.

 

B. The Foundations of Copyright Law in Malaysia

To evaluate how Malaysian copyright law may respond to the challenges posed by generative AI, it is essential to first understand the legal structure, scope, and principles underpinning copyright protection in Malaysia. The Copyright Act 1987 (‘the Act’), supplemented by various subsidiary legislation, forms the cornerstone of Malaysia’s copyright regime. Unlike other branches of intellectual property law such as patents or trademarks copyright protection arises automatically upon the creation of an eligible work. No formal registration is required.

This section outlines the categories of protected works, key statutory requirements, the distinction between ideas and expression, and the dual nature of copyright rights in Malaysia.

B1. What Can Be Copyrighted?

Under Section 7(1) of the Act, copyright subsists in specific categories of original works, namely:

• • Literary works
  • Musical works
  • Artistic works
  • Films
  • Sound recordings
  • Broadcasts

In addition to these original works, Section 8 extends protection to derivative works, such as:

• • Translations, adaptations, arrangements and other transformations of works eligible for copyright; and
  • Compilations or collections of data that involve intellectual creation in their selection or arrangement.

For example, a curated travel guidebook with thematic organisation and commentary may be protected, whereas a raw, alphabetical list of legal case names would not qualify, as it lacks originality or creative arrangement.

 

B2. Copyright Protects Expression, Not Ideas

A central tenet of copyright law is that only the expression of an idea is protected and not the idea itself. This is codified in Section 7(2A) of the Act, which excludes from protection any idea, procedure, method of operation or mathematical concept as such.

This section ensures that copyright does not monopolise general themes, facts, or creative concepts. For example, while a particular love song about heartbreak such as its lyrics, melody, and arrangement may be protected, the broader theme of heartbreak in music remains free for anyone to use.

B3. Infringing Works Can Still Qualify for Copyright?  

Interestingly, Section 7(4) provides that a work is not automatically disqualified from copyright protection merely because it was created by infringing someone else’s copyrighted work. In other words, even an infringing work can attract copyright, so long as it meets the statutory requirements. This reflects the automatic and unexamined nature of copyright protection in Malaysia, that is, once a qualifying work is created, protection arises. However, this does not excuse the infringement. The original copyright owner retains the right to enforce their rights and may still initiate legal action.

This stands in stark contrast to other areas of intellectual property law:

• 1. In patent law, if your invention infringes an existing patent, you cannot obtain your own patent.
  2. In trademark law, confusingly similar marks are refused registration or invalidated.

This anomaly in copyright law reflects its automatic nature, but also introduces complexity, especially in the context of AI-generated or derivative works, where the line between influence and infringement may be blurred.

 

B4. Requirements for Copyright Protection

Copyright protection is a means to prevent others from taking unfair advantage of works which they did not create. For a work to qualify for protection under Malaysian law, it must meet 4 statutory conditions, as follows:

1. 1. Originality

The work must be original. This, however, does not mean that the work must be new or innovative (see Kiwi Brands (M) Sdn Bhd v Multiview Enterprises Sdn Bhd [1998] 6 MLJ 38 at p 46), but it must be created with sufficient labour and skill expended in creating the work as per section 7(3) of the Act. This is a low threshold. The phrase “sufficient effort” in section 7(3) is inextricably linked to the degree of “effort, skill or labour” expended in the creation of the work and even though the word “or” is there, both the elements of skill and labour must be proven: Megnaway Enterprise Sdn Bhd v Soon Lian Hock (sole proprietor of the firm Performancee Audio & Car Accessories Enterprise [2009] 3 MLJ 525 at [26].

1. 2. Fixation in Material Form

Section 7(3)(b) stipulates that the work must be written down, recorded, or reduced to material form. There must be a physical embodiment or fixture of the expressions of your ideas. One of the reasons is that this ensures that the expression is available for others to observe and determine whether their work is potentially infringing. This applies only to literary, musical or artistic work.

However, for other categories of work, there must be some degree of permanency in fixation required. Digital storage (e.g., on a USB drive) can suffice.

1. 3. Satisfying One of the 3 Qualifications under Section 10

A work is eligible for copyright protection in Malaysia if it satisfies any one of the following:

(a) The author is a qualified person;

Under Section 10(1)(a), a work qualifies if, at the time it is made, the author is a qualified person, as defined in Section 3:

1. 1. For individuals: a citizen or permanent resident of Malaysia; and
   2. For corporate entities: means a body corporate established in Malaysia and constituted or vested with legal personality under the laws of Malaysia.

(b) The work is first published in Malaysia;

Even if the author is not a Malaysian citizen or resident, Section 10(1)(b) grants protection if the work is first published in Malaysia. Section 4(3) further clarifies the meaning of first publication:

• 1. The work was first published in Malaysia and not elsewhere, or
  2. The work was first published elsewhere but published in Malaysia within 30 days of that initial publication.

This means that foreign authors can still obtain copyright protection in Malaysia if they publish in Malaysia within the relevant time frame, thereby extending protection to international creators and encouraging creative dissemination within the country.

(c) The work is made in Malaysia

Lastly, under Section 10(1)(c), a work qualifies for protection if it is made in Malaysia. In Hexagon Tower Sdn Bhd v Polydamic Holdings Sdn Bhd & Ors [2005] CLJU 77, the court held that a work is “made” when it is completed, not merely started. This provision captures works created within the Malaysian jurisdiction, regardless of the nationality of the author.

1. 4. The work must fall within the categories outlined in Sections 7 or 8 of the Act mentioned above.

B5. Authorship and Ownership

The concept of authorship lies at the heart of copyright law. Section 3 of the Act defines the “author” based on the type of work. For instance, a writer for literary works, a composer for music, or the person arranging for the creation of a film or photograph. Importantly, authorship and ownership may diverge: while the author is the creator of the work, ownership of copyright can be assigned to another party.

With this legal foundation established, the article will now turn to the copyright challenges arising from generative AI, beginning with an exploration of authorship and originality in an era where machines—not humans—are creating content.

C. AI and Emerging Copyright Challenges

It is unsurprising that the rise of generative AI has opened doors to new forms of creativity, sparking both excitement and uncertainty within the legal world. While AI tools have empowered individuals—from artists and writers to marketers and developers—to create content faster and in new, imaginative ways, they have also thrown a wrench into how we understand authorship, originality, and ownership under copyright law.

Traditionally, copyright law is built around the idea of human creativity—protecting the fruits of one’s intellect, imagination, and labour. But when machines start generating poems, paintings, essays, and even entire films, often with minimal human input, the core principles of copyright are tested. Who owns the output? Can a machine be an “author”? And what happens when these AI systems learn by consuming billions of pieces of existing (and copyrighted) content?

Although Malaysia has not yet witnessed legal disputes involving generative AI, several other jurisdictions are already grappling with the complex challenges AI presents to copyright law. Among the most commonly contested issues are the following:

C1. The Problem of Training Data

To function effectively, AI systems especially generative models like ChatGPT, Midjourney, or Stable Diffusion rely on training data. Training data includes massive datasets collected from books, websites, artwork, music, and various forms of media. Here lies the problem: a significant portion of this content may be protected by copyright. Many AI developers have scraped publicly available data without always seeking permission or paying royalties. On the surface, it might seem harmless but when that learning includes in-depth analysis and replication of creative works, it becomes a legal minefield.

In Silverman v. OpenAI (2023), a group of published authors alleged that OpenAI trained its language models (including ChatGPT) by scraping datasets containing their copyrighted books without obtaining permission or providing compensation.

In Andersen & Ors v. Stability AI (2023), a group of visual artists filed a lawsuit and argued that image-generating AI models such as Stable Diffusion, DreamUp (by DeviantArt), and Midjourney had been trained on millions of copyrighted images including theirs without consent, credit, or compensation.

These cases underscore a growing tension: should using copyrighted works to train AI be considered fair use, or is it outright infringement?

C2. Originality and the Question of Authorship

One of the foundational pillars of copyright law is originality, meaning a work must result from some level of personal creativity or intellectual effort. But with AI-generated content, originality gets murky.

Consider this: if you type a prompt like “paint a serene mountain landscape at sunrise in the style of Studio Ghibli” into an AI image generator, and it produces a stunning visual, who exactly is the creator? Is it you, for crafting the prompt? The AI company, for building the system? Or the machine itself? Most legal systems agree on one thing: machines cannot be authors. Copyright law still requires a human touch. The question lies then on how much human input is enough? Several jurisdictions have considered this question.

In the United States, the case of Théâtre D’opéra Spatial garnered significant attention as one of the earliest legal flashpoints involving AI-generated art. Jason Allen, a game designer, used the generative AI tool Midjourney to create a digital artwork, which he later entered into the Colorado State Fair’s art competition in 2022 where it controversially won first place in the digital art category. Allen emphasized that he had put substantial effort into the creation, including the use of over 600 uniquely crafted prompts, followed by refinement, upscaling, and adjustments using additional tools like Photoshop. Despite this human involvement, the U.S. Copyright Office ultimately denied his application for copyright registration. In its final determination on 5 September 2023, the U.S. Copyright Office concluded that the human contribution in the piece was de minimis as the majority of the creative expression originated from Midjourney’s autonomous generation process. The decision reaffirmed the U.S. Copyright Office’s position that copyright protection in the U.S. is reserved strictly for works of human authorship, and that significant creative control beyond merely inputting prompts is necessary.

Similarly, Zarya of the Dawn, a comic book created by author Kristina Kashtanova, sparked significant attention in 2022. Kashtanova used Midjourney to generate the comic’s illustrations while writing the text and planning the narrative herself. Initially, the U.S. Copyright Office granted registration but later partially rescinded it in March 2023, stating that while the text and arrangement were eligible for copyright, the AI-generated images were not. The U.S. Copyright Office stressed that Kashtanova had not exercised sufficient control over the visual output to claim authorship of those images under U.S. copyright law. This case further solidified the U.S. Copyright Office’s stance: AI-generated content, absent meaningful human intervention, is not protectable, but structured human contributions like text, layout, and narrative may still receive protection.

In contrast, consider the recent case of A Single Piece of American Cheese, where the U.S. Copyright Office granted partial copyright protection—not for the individual AI-generated elements themselves, but for the selection, coordination, and arrangement of those elements by a human author. The applicant had used an AI image generator to create multiple visual segments and then manually curated, cropped, positioned, and combined those segments into a composite image; see https://youtu.be/H4gIIjUGOhY).

While the U.S. Copyright Office made it clear that the individual AI-generated parts were not eligible for protection due to the lack of human authorship, it acknowledged that the creative manner in which the segments were assembled reflected sufficient human originality. This case underscores that human creativity exercised in the curation, modification, and composition of AI-generated materials can still qualify for copyright protection provided that the resulting work demonstrates independent intellectual effort and expressive choices.

In the landmark Chinese case of Li v. Liu (2023), the Beijing Internet Court became one of the first courts globally to recognize copyright protection for an AI-generated image. The plaintiff, Mr. Li, used the generative AI model Stable Diffusion to create a digital illustration, which he then published on the Xiaohongshu (Little Red Book) platform. The defendant, Mr. Liu, subsequently used this AI-generated image in an article published on Baidu’s Baijiahao platform without permission, prompting Mr. Li to file a lawsuit for copyright infringement.

The court ruled in Mr. Li’s favor, finding that the image was indeed a copyrightable work under Chinese law. The decision was based on several key findings:

1. 1. Human Intellectual Contribution: The court emphasized that Mr. Li had demonstrated a clear and substantial intellectual effort. He had designed the image by formulating detailed prompt words, adjusted parameters within the AI system, arranged the order of inputs, and repeatedly refined the outputs until selecting a final version that matched his intended creative vision.

1. 2. Originality: The court held that the resulting image satisfied the requirement of originality, as it reflected personalized expression arising from Mr. Li’s individual aesthetic choices and control over the AI process. It was not merely a product of random generation.

1. 3. “Intellectual Achievement” Requirement: Under Chinese copyright law, a protected work must embody an “intellectual achievement.” The court reasoned that Mr. Li’s input constituted such achievement, as it involved intentional, creative manipulation of the generative process, not mere passive prompting.

This ruling marked a significant departure from the approach taken in jurisdictions like the United States, where AI-generated works lacking direct human authorship are generally deemed ineligible for copyright. In contrast, the Beijing court affirmed that AI can serve as a tool, and when used under human direction with creative intent, the output may be protected as the result of human authorship.

D. Comparative Analysis from Different Jurisdictions

D1. The Position in China

As enunciated in Li v. Liu (2023) and Shenzhen Tencent v. Shanghai Yingxun (2019), the Courts in China are generally willing to recognise copyright protection for AI generated works provided that substantial human involvement is established on a case-by-case basis.

In Shenzhen Tencent (2019), Tencent Shenzhen used Dreamwriter (a writing assistance system based on data and algorithm) to generate an article published on the Tencent finance website. The Shenzhen Nanshan District Court held that the article was copyrightable as the preparatory work (i.e. data input, trigger condition creating, template and corpus setting) done by the team were considered intellectual activities directly related to the article generated subsequently. Further, although Dreamwriter generated the article within two minutes, it required human involvement. The operating process also reflected the choices and decisions made by the team who generated the article. In this regard, China’s stance with regard to affording copyright protection to AI generated works could encourage innovation and promote the use of AI.

 

D2. The Position in the United States

In the United States, a strict human authorship position (i.e., works created by a human being) is adopted for works to be eligible for copyright protection. Therefore, fully autonomous AI-generated content is excluded.

The case of Thaler v. Perlmutter (2023) revolved around whether the work authored by a computer system could be eligible for copyright, and if yes, whether the copyright could be transferred to the owner of the software. The District Court for the District of Columbia held that, in the absence of any human involvement in the creation of the work, the work generated autonomously by a computer system is not eligible for copyright, thus reaffirming the human authorship requirement in 17 U.S. Code 102. Given that the work at issue did not give rise to a valid copyright upon its creation, the issue of how ownership of such a copyright could have passed to the plaintiff in the case was not further addressed.

Further, in Zarya of the Dawn (2023), as explained in Section C2 above, when deciding whether the individual images are copyrightable, the U.S. Copyright Office considered Ms. Kashtanova’s use of Midjourney’s AI technology in its copyrightability analysis. The U.S. Copyright Office sought to understand how Midjourney functions by offering an AI technology capable of generating images in response to text provided by a user. Users operate Midjourney through “prompts,” which are text commands entered in one of Midjourney’s channels.  Users also have the option to include (i) a URL of one or more images to influence the generated output; or (ii) parameters directing Midjourney to generate an image in a particular aspect ratio or providing other functional directions. The initial prompt by a user generates 4 different images based on Midjourney’s training data.

The U.S. Copyright Office concluded that Midjourney does not interpret prompts as specific instructions to create a particular expressive result. While additional prompts applied to one of these initial images can influence the subsequent images, the process is not controlled by the user because it is not possible to predict what Midjourney will create ahead of time. Therefore, the U.S. Copyright Office concluded that the images generated by Midjourney, as contained within the work, are not original works of authorship protected by copyright due to the lack of predictability of the output.

Unlike the Chinese Courts, which are willing to recognise works generated via AI with the use of the users’ prompts, the U.S. Copyright Office took a more conservative approach and examined both the nature of the AI system and the degree of user influence over the final outcome. If Zarya of Dawn (2023) had been decided by the Chinese Courts, it is likely that the images produced by Ms. Kashtanova via the use of Midjourney would have been eligible for copyright protection as Ms Kashtanova did use text prompts and other parameters when she generated the images in question.

Based on the U.S. Copyright Office’s letter to Ms Kashtanova, the outcome may have been different if the images in question were produced by an AI assisted tool, in which the users are able to actively guide and influence the creative process. However, the issue of how much “control” would satisfy the human authorship requirement remains uncertain. Given the increasing use of AI tools in image and text production, it may be timely to revise the copyright laws in the United States to clarify the level of human authorship or creative influence required to be eligible for copyright protection.

 

E. The Current Policy Landscape on AI in Malaysia

While other jurisdictions have begun testing the boundaries of copyright law through litigation and administrative decisions, Malaysia remains in the early stages of addressing the intersection of AI and intellectual property. Although no legal disputes concerning generative AI have yet arisen locally, Malaysia has started laying the groundwork for responsible AI development through policy initiatives. However, key legal questions, particularly those concerning authorship and ownership of AI-generated works, remain unanswered.

The following section outlines the current policy environment and legal landscape surrounding AI in Malaysia.

E1. Soft Law

Malaysia has taken early steps toward formalising the governance of AI, including generative AI, with the establishment of the National Artificial Intelligence Office (‘NAIO’) in December 2024 and the release of the National Guidelines on AI Governance and Ethics (‘AIGE’) by the Ministry of Science, Technology and Innovation (‘MOSTI’) in September 2024.

The AIGE serves as a non-binding policy framework to promote responsible AI practices, particularly through the lens of ethics, safety, and societal impact. It targets three core stakeholder groups—society, government agencies and policymakers, and industry—and is aligned with international standards, particularly those issued by UNESCO.

AIGE is built around seven key principles:

• • Fairness
  • Privacy and Security
  • Pursuit of Human Benefit and Happiness
  • Inclusiveness
  • Accountability
  • Transparency
  • Reliability, Safety, and Control

However, the AIGE currently only offers broad guidance and lacks legal enforceability. Furthermore, it does not directly address the legal implications of generative AI, particularly in the domain of intellectual property. Where copyright law is concerned, the guidelines merely acknowledge that infringement risks exist, with a note that these will be considered more substantively in a future version of AIGE that will focus on generative AI.

E2. Case Law

At present, Malaysia has no case law addressing the copyright implications of generative AI.  This legal gap creates uncertainty for creators, platform operators, and rights holders navigating a rapidly evolving landscape where AI-generated content often does not fit neatly within traditional copyright frameworks.

The Act does not expressly define “author” to include or exclude non-human creators. However, Section 17(1), which sets out the duration of copyright for literary, musical, and artistic works, states that protection lasts “during the life of the author and for 50 years after his death.” This wording presumes a human author, as concepts like lifespan and death do not apply to AI systems. Accordingly, it supports the view that Malaysian copyright law, at least under current interpretation, recognises only human authorship.

However, this does not conclusively resolve the issue of whether a human user of an AI system, for instance someone who inputs prompts, can be considered the author of the resulting output. That would depend on whether the user’s involvement meets the threshold of originality and sufficient effort, as required under Malaysian law. These are fact-specific inquiries, and without judicial guidance, the legal position remains unsettled.

For now, it remains uncertain whether generative AI-generated works, particularly those produced with minimal or no human intervention, can attract copyright protection in Malaysia. In the absence of binding precedent or legislative reform, creators and users of such content operate in a legal grey area.

 

F. Future Outlook: Regulating Generative AI and Addressing Legal Uncertainty

Looking ahead, Malaysia must consider how best to adapt its legal framework to address the evolving challenges posed by generative AI. Two key questions arise:

1. Should Malaysia introduce a standalone AI Act (as seen in jurisdictions such as the European Union and China) or instead amend existing legislation to accommodate AI-generated works?

2. How should courts approach the issue of authorship, particularly in situations where human input in the creation of AI-generated content is minimal or indirect?

Given the rapid and unpredictable development of AI technologies, a standalone AI Act may risk becoming outdated or overly general before it can meaningfully address specific legal issues. A more flexible and targeted approach may be to amend existing legislation, such as the Act, and allow case law to gradually shape the legal framework. Case-by-case judicial interpretation offers the advantage of being responsive to novel fact patterns and better suited to addressing real-world use cases as they arise.

On the second question, Malaysia will eventually have to determine whether to follow the United States’ approach, which strongly emphasises human creativity as a prerequisite for copyright protection, or consider a China-style approach, which has shown some willingness to recognise copyright in AI-generated works where human intention and contribution can be demonstrated.

In this context, it is instructive to consider the decision by the U.S. Copyright Office involving ‘A Single Piece of American Cheese’, where partial copyright protection was granted. The U.S. Copyright Office accepted protection for the selection, coordination, and arrangement of visual elements made by a human user, despite the involvement of AI. Importantly, the human author in that case had used an interactive or mixed-initiative generative AI tool, akin to Adobe Firefly in Photoshop, which allows users not only to prompt the AI but also to manipulate, refine, and override the generated content. These tools afford users a high degree of creative control, positioning them more clearly as the authors of the final work.

This approach appears to strike a reasonable balance. It acknowledges that users may generate protectable works using generative AI, without abandoning the foundational principle that copyright exists to reward human creativity and intellectual effort. Moving forward, Malaysia may consider adopting a similar framework of recognising copyright where there is clear human authorship, while continuing to deny protection for outputs that are fully machine-generated with random results that have no meaningful human input.

 

G. Conclusion

Generative AI is reshaping how creative works are produced, raising complex questions for copyright law in Malaysia and beyond. As AI systems become more sophisticated, determining whether, and when, human users can claim authorship over AI-generated content is becoming increasingly urgent.

Malaysia has made initial strides in AI governance through initiatives like the NAIO and the AIGE but these soft law measures fall short of resolving the legal uncertainty surrounding generative AI and copyright. Unlike jurisdictions such as China or the United States, Malaysia has yet to see decisive case law or targeted legislative reform in this area.

Moving forward, Malaysia may benefit from a pragmatic, case-by-case approach. Rather than introducing sweeping new laws, targeted amendments to the Act, supported by judicial interpretation, could provide the flexibility needed to address the evolving role of AI in creative industries. In doing so, Malaysia will have to decide whether to adopt a strict human authorship model, as seen in the United States, or to take a more inclusive stance like China, where human intention and creative input can still attract protection, even in AI-assisted works.

Striking the right balance is key. Copyright must continue to incentivise genuine human creativity while recognising the realities of AI-assisted production. A clear, adaptable legal framework will be essential to ensuring that Malaysia remains ready to meet the challenges—and opportunities—of generative AI in the digital age.