I. INTRODUCTION

The rapid development of artificial intelligence (‘AI’) technologies is transforming the landscape of creative production, paving the way for significant questions regarding intellectual property rights for AI-generated content. As AI systems increasingly produce music, art, literature, and other forms of creative works, the issues surrounding the relevance of copyright protection for these works become more critical. It may be argued that extending copyright to AI creations recognises the innovative potential of these technologies and raises more investment in their growth. On the contrary, critics stressed that granting copyright to AI-generated works may weaken traditional fundamentals of copyright and ownership, resulting in complex legal challenges and ethical dilemmas. This paper seeks to provide an overview of the different jurisdictions of granting copyright to AI-generated works.

II. BACKGROUND

The term ‘artificial intelligence’ encompasses several meanings depending on its interpretation and context. Hossain, Miraz and Ya’u propose that AI refers to computer programs capable of performing complex tasks that were once exclusive to human abilities, such as problem-solving, reasoning, and decision-making.[1] In other words, AI is a technology that allows computers or computer systems to perform tasks that typically require human intelligence. AI has progressively expanded its performance as the system is being used in multiple areas including healthcare, finance, technology, and business.[2]

AI technology first gained widespread public recognition in November 2022, when OpenAI introduced ChatGPT to the public.[3] This release prompted people to adopt AI as a tool to complete their tasks more quickly and effectively. As ChatGPT reached higher levels of public awareness, other developers began launching their own AI applications with similar capabilities, attracting additional public interest. However, as AI’s abilities to fully replicate human work were still limited at that time, copyright issues were not a prominent concern. It was only by early 2024, when roughly twenty lawsuits had been filed concerning copyright infringement by the AI tools’ developers, that these issues came to the forefront.[4] By that time, generative AI was already invented and popularly used to generate pieces like artworks, literary texts and even song covers.[5] This is when the legal facets of AI’s work, especially in terms of copyright, become critically important.

The main question surrounding AI-created works is: who should be entitled to copyright? Should it be the AI itself, or the developers of the AI, or the users of the AI? These seemingly simple questions continue to provoke debate in the legal world. Guadamuz argues that copyright for AI-generated works has never been expressly prohibited by the law. Instead, most jurisdictions limit copyright protection to only human-made works.[6]

As artificial intelligence continues to evolve, it also presents new opportunities and challenges, particularly in the realm of intellectual property (‘IP’) rights. The capacity of AI to create original works, from visual art and music to written content, has driven debates about the consequences of copyright protection for these AI-generated outputs.[7]

This intricacy necessitates a careful examination of existing legal frameworks and the potential need for reforms to address the unique challenges posed by AI in creative domains. However, recent case decisions demonstrate that the implementation of these existing legal frameworks may not be as clear-cut, and different cases can result in diverse outcomes.

III. INTERNATIONAL PERSPECTIVES

Copyright law serves the purpose of protecting the original work of authorship.[8] When AI developed the ability to create human-like workpieces such as artworks and literary writings, scholars have argued and provided different stances on whether AI should be granted legal authorship.[9] This issue has become more relevant today as legal frameworks interpret and define authorship distinctively across varying contexts.[10] For instance, certain legal frameworks uphold that authorship is a fundamentally human endeavour, as highlighted by cases such as Feist Publications, Inc., v Rural Telephone Service Co.[11] in America, where originality is a key principle for copyright. However, as AI systems create works mimicking human originality, countries differ on how or if copyright applies to non-human authors.

1. The European Union

The European Union (‘the EU’) emphasises human authorship,[12] focusing on human creation as the basis for copyright eligibility. This stance stems from the EU’s principle that copyright protection is rooted in human creativity and expression, thus excluding non-human creators such as AI from being eligible for copyright on their own. This can be demonstrated in the landmark case of Infopaq v Danske Dagblades Forening where copyright protection applies to any work that originates from the author’s unique intellectual efforts.[13] In addition, the court in Eva-Maria Painer v Standard VerlagsGmbH and Others elaborated that an intellectual work must reflect the author’s personality, express their free and creative choices, and must have a personal touch.[14] Consequently, AI cannot be considered authors under EU copyright law as they are incapable of independent creative expression.

A recent case in the Czech Republic has further examined the issue of copyright protection for AI-generated works. In 2023, the Municipal Court in Prague heard a case concerning an image generated using AI.[15] The plaintiff had generated an image using the prompt: ‘create a visual representation of two parties signing a business contract in a formal setting, such as a conference room or a law firm office in Prague. Just show your hands.’ The plaintiff later sued the defendant for publishing the said image on their website without the plaintiff’s consent, seeking injunctive relief.

The court adjudicated two key issues: (i) whether an AI-generated image can be granted copyright protection; and (ii) whether the natural person who created the prompt that led AI to generate the image can be considered as its ‘author’. In its judgement, the court held that the AI could not be considered the ‘author’ as only a natural person can be the author of a copyrighted work, aligning with the EU’s stance. Furthermore, a work of authorship must be the unique result of the creative activity of a natural person. While the plaintiff provided the text prompt, the court found that this alone did not constitute sufficient creative authorship, as the work was not created personally but with the help of artificial intelligence. The court further held that the prompt, at most, is the ‘subject matter of the work’ or an ‘idea’ but are not in themselves works of authorship deserving protection under Czech law. Since the plaintiff could not prove authorship, they lacked standing to claim copyright infringement, and the case was dismissed.

It should be noted however, that the court dismissed the case largely due to the lack of evidence provided by the plaintiff in proving that their prompt directly resulted in the AI-generated image. The judgement leaves open the possibility that AI-generated works could be granted protection if the prompt used to generate it sufficiently demonstrated the author’s originality.[16] Whether this reasoning will be adopted in other EU member states remains to be seen.

2. The United Kingdom (‘the UK’)

The UK’s Copyright, Designs and Patents Act 1988 (‘CDPA’) includes a unique provision for computer-generated works under Section 9(3), which stipulates that the ‘author’ of a computer-generated work is the person who undertakes the necessary arrangements for the creation of the work (such as programming).[17] This essentially allows for copyright protection despite the absence of direct human authorship.[18] Nonetheless, challenges to the provision’s applicability remain, some of which include establishing originality, disclosing AI involvement, and seeking a human author, all of which hinder the effectiveness of granting copyright protection for AI-generated works under the section.[19]

Among these challenges, the most significant is the requirement for a computer-generated work to be ‘original’ before it can be protected.[20] The standard for originality in the UK has evolved, with earlier cases requiring the author to have exercised their own skill, labour, and judgment,[21] while increasingly, more recent legal decisions, particularly those under EU influence, demand that a work be ‘the author’s own intellectual creation’.[22] In this context, originality must stem from the author’s free and creative choices rather than a mechanical process.[23] Therefore, despite Section 9(3) sounding flexible to computer-generated works, reality often reflects otherwise as a copyrighted work must still involve human elements and creativity.

Some scholars suggest that the contention under Section 9(3) should therefore adopt a more nuanced approach. One such argument is that under Section 9(3), AI-generated works should be owned by the user who provides instructions to the AI model, but only if those instructions are sufficiently original to qualify for copyright protection.[24]

The question of copyright protection for AI-generated works also raise broader concerns about AI’s role in intellectual property, where similar questions arise about whether AI can hold ownership rights. A pertinent issue was addressed in Thaler v Comptroller-General of Patents, Designs and Trade Marks,[25] which focused on the issue of whether AI-based machines can make patentable inventions. The appellant of this case is the owner and creator of an AI machine named DABUS. He filed two patent applications designating DABUS as the inventor. However, the Supreme Court held that DABUS is legally incapable of being an inventor under UK Patent Law because an inventor must be a natural person under the Patents Act 1977. The court further held that the appellant is not entitled to apply for patents as the owner of DABUS. His argument based on the doctrine of accession was rejected, as the court found no legal basis for treating an AI-generated invention as tangible property that automatically transfers to the owner of the AI system. The doctrine also does not apply to patents as a matter of law. Based on this case, it can be concluded that neither the owner of the AI machine nor the AI machine itself has the right to be granted ownership for a work produced by a machine.

Coming back to the discussion of Section 9(3), it has been demonstrated that a natural human is still a pre-requisite for ownership although the wording of the section seems to give leeway for computer-generated works to be granted copyright eligibility. Therefore, it is submitted that the UK remains reliant on human ownership over AI ownership.

3. The United States (‘the US’)

Unlike the UK, the copyright regime in the US traditionally requires human authorship to qualify for copyright protection.[26] The US Copyright Office’s stance is illuminating: ‘If a work’s traditional elements of authorship were produced by a machine, the work lacks human authorship and the Office will not register it’.[27] Accordingly, if a human user’s sole contribution was to instruct an AI system to generate material, the ‘traditional elements of authorship’ – such as expressive elements – are determined and executed by the AI, not the human user.[28]

However, AI-generated material may be copyrightable if a human user adds sufficient human authorship — for example, by arranging or selecting AI-generated content, or significantly modifying the material.[29] In such cases, only the human-authored aspects are protected, not the AI-generated content.[30] It follows that works produced autonomously by a machine without any human input would not qualify for copyright protection.[31]

The issue of AI authorship in copyright law was directly addressed in Thaler v Perlmutter[32] where the US District Court for the District of Columbia upheld the US Copyright Office’s decision to deny copyright for an artwork generated entirely by AI without any human input. The court reiterated that copyright law requires human authorship, aligning with the US Copyright Act 1976.

Another interesting copyright decision to discuss is Zarya of the Dawn,[33] decided in 2023. It involves an artist who used AI to generate images for their comic book. The USCO partially granted copyright registration for the human-authored text of the work, and the selection, coordination and arrangement of the images and text, as they had the sufficient human creativity required to qualify for protection.[34] However, despite providing ‘hundreds or thousands of descriptive prompts’ to the AI system,[35] the US Copyright Office (‘USCO’) ruled that the AI-generated images were not eligible for copyright protection. The USCO found that the distance between what a user instructs the AI system to create and the material it actually produces was too significant for the artist to have sufficient control over the images. Unlike a camera, for example, AI operates unpredictably, making it difficult for users to have comparable creative control over the final result. The USCO also likened the situation to that of a client who directs an artist to create an image. Despite the numerous directions given (and absent the legal requirements for a work made for hire), the visual artist who received those instructions and determined how best to express them is the author.[36]

The question of authorship in Zarya of the Dawn mirrors a similar debate in patent law: whether AI can be recognised as an inventor. In 2022, the United States Court of Appeals for the Federal Circuit heard the case of Thaler v Vidal[37] where the legal question and reason for appeal are alike: who, or what, can be an inventor? Specifically, the court was asked to determine whether an AI software system could be listed as an ‘inventor’ on a patent application. The Federal Circuit denied the appellant’s claim, holding that the US Patent Act requires that inventors must be natural persons; that is, human beings.

The appellant also argued that the court’s refusal to recognise AI as an inventor would hinder the constitutional purpose of patents to promote the progression of science and the useful arts. However, the court rejected this argument, holding that the constitutional clause the appellant cited is merely a grant of legislative power to Congress. Since Congress has chosen to exercise that power by passing the Patent Act (which limits inventorship to human beings), his argument was deemed inapplicable. The court reaffirmed that only natural persons can be inventors under the Patent Act.

Not much has changed with regard to the US’s stance in 2025. The USCO, in a report published in January 2025, reaffirmed that copyright protection is contingent on human authorship but acknowledged that AI-generated works could be protected if a human author contributes sufficient originality.[38] The current legal standard in the US maintains copyright protection is limited to works with human authorship. However, recent developments suggest a way for a more flexible approach, allowing protection for AI-generated works if there is sufficient human input.

4. Australia

For literary, dramatic, musical and artistic works in Australia, the work must be created by a human author[39] who contributed ‘independent intellectual effort’.[40] As AI systems are neither human persons nor Australian citizens, they are not entitled to authorship under the Act. However, AI-assisted works by human authors may enjoy protection – provided that there is sufficient independent intellectual effort.

Unfortunately, at the time of writing, there is a dearth of Australian legislation or case law that directly addresses this issue. In such an event, an examination of relevant cases may offer insight into the courts’ likely approach. One such case is IceTV Pty Ltd v Nine Network Australia Pty Ltd.[41] It involves Nine Network, a major broadcaster, which sued IceTV for allegedly reproducing its weekly programme schedules for an electronic programme guide known as ‘IceGuide’. The alleged parts copied were the time and title information of scheduled television shows. The key legal question here was whether Nine Network’s television schedules were sufficiently original to deserve protection under the Copyright Act 1968.

The High Court ultimately ruled in favor of IceTV, holding that copyright protects original creative expression, not mere facts or information. It found that while IceGuide incorporated factual elements such as time and title information, these elements were not protected by copyright. The court found that the originality of Nine Network’s weekly schedule lay in the selection and presentation of that information together with additional program details and synopses. As a result, the court ruled that IceTV had not infringed Nine Network’s copyright, as IceTV did not reproduce a substantial part of Nine’s original creative expression but only its unprotected factual data. Therefore, applying the court’s reasoning to the present discussion, AI-generated content may not qualify for copyright protection unless a human author contributes creative intellectual effort to the work, such as modifying or curating the output.

Due to limited case law on AI-generated works in copyright, it might be pertinent to examine cases where courts have grappled with AI’s role in intellectual property. One such case that took a more liberal stance on AI-created works is the 2021 Federal Court case of Commisioner of Patents v Thaler[42] in Australia. Similar to the aforementioned American case of Thaler v Vidal, the court was asked whether or not an ‘inventor’ under the Patents Act 1990 and the Patent Regulations 1991 can be an AI system. Justice Beach reasoned that an AI system could be regarded as an inventor under the Patents Act. He argued that the term ‘inventor’ is an agent noun, which implies that it can refer to either a person or an entity responsible for inventing. This interpretation aligns with practical circumstances, given that in many cases, certain inventions cannot reasonably be attributed to a human inventor alone. Furthermore, he observed that nothing within the text of the Patents Act contradicts this interpretation, thus supporting the possibility for an AI to be recognised as an inventor under current patent laws.

However, this decision was ultimately overturned by the Full Court of the Federal Court of Australia in Thaler v Commissioner of Patents,[43] which ruled that the Patents Act implicitly requires human inventorship. Nonetheless, the earlier ruling reflects a potential shift in some judicial circles toward a more positive stance of AI-generated works.

Recognising growing challenges due to the absence of copyright protection law for AI-generated works in Australia, the Australian government has acknowledged the need for further examination. An interim response published by the Australian government in 2023 highlighted ongoing research and consultation by government bodies to assess the effects of AI on copyright and broader IP law.[44]

5. China

China has a unique perspective in this discussion. Although the Copyright Law of the People’s Republic of China did not expressly acknowledge AI-created works, recent cases have developed an intriguing insight.[45]

A notable example that sides with AI-generated works can be seen in Shenzhen Tencent Computer System Co Ltd v Shanghai Yingxun Technology Co Ltd,[46] also known as the Dreamwriter case. Here, the court admits the novelty of AI works and takes a positive stance towards AI-created pieces.[47] The case involves a writing robot known as ‘Dreamwriter’, developed by Tencent. Tencent subsequently sued another company for copying and publishing an article written by Dreamwriter on their website, alleging copyright infringement.

Instead of granting authorship to Dreamwriter, the Nanshan District People’s Court granted authorship to its human developers. The court found that the creation process of an article generated by Dreamwriter involved human intellectual effort, including providing data and selecting templates for the AI system, and making creative decisions about the article’s structure, such as its theme and writing style. Therefore, the court ruled that, while Dreamwriter carried out the actual writing, the form of expression in the article was a result of the team’s personalised choices and arrangements made by its developers.

Despite this, it still falls short of being a landmark ruling that grants copyright to AI itself. A closer examination of the verdict reveals that it did not break away from existing copyright law or acknowledge AI as an author. As Tencent's creative team played a crucial role in producing the article, Dreamwriter functioned merely as an intelligent tool assisting Tencent, which was ultimately recognised as the rightful author.[48]

The Li v Liu[49] case also marked a significant development in China’s approach to AI-generated works. Unlike the USCO’s decision in Zarya of the Dawn, the Beijing Internet Court ruled in favour of protecting AI-generated images under copyright law, and that the plaintiff who prompted the AI was entitled to authorship under Chinese copyright law. Here, the plaintiff generated an image of a young woman and published it on Xiaohongshu, a Chinese social media platform. When the image was subsequently used by the defendant without the plaintiff’s consent, the plaintiff sued for copyright infringement.

For a work to be considered copyrightable under Chinese law, it must, inter alia, be original, and constitute an intellectual achievement.[50] ‘Intellectual achievements’ as defined by the court, refers to ‘the results of intellectual activities’, meaning it must ‘reflect the intellectual input of a natural person’. In reaching the final image, the plaintiff used specific prompts, set parameters to modify its composition and layout, and refined each generated image until he achieved his desired result. The court recognised the plaintiff’s selection and arrangement of AI-generated elements as sufficient intellectual input, satisfying the requirement of ‘intellectual achievement’.

The court also held that the image was ‘original’. The standard of originality, as expounded by the court, requires that the ‘work be completed independently by the author and reflect the author’s personalised expression’. Hence, while the court acknowledged that AI generated the images, it found that the human’s creative intent and selection process contributed to their originality. The human user’s iterative prompting and changing of parameters until he obtained his final result demonstrate it wasn’t mere mechanical intellectual achievement, but instead reflected his personalised expression. The court concluded that the plaintiff who generated the image, not the AI or its developers, was the rightful author of the image, and found the defendant guilty of copyright infringement. This case therefore is a step forward in acknowledging AI-assisted works as eligible for copyright protection under China law.

IV. CONCLUSION

In conclusion, the recognition and protection of intellectual property rights for AI-generated works presents a complex, evolving challenge across international legal frameworks. Copyright protection for AI-generated works varies significantly across legal systems, as individual countries have established distinct criteria for determining authorship in algorithmic creations. This diversity stems from the complexity of recognising non-human creators within traditional frameworks of originality. Consequently, these varied legal interpretations contribute to global inconsistencies regarding whether AI-generated works should be eligible for copyright. While some jurisdictions, such as China, have taken progressive steps in interpreting human authorship by granting AI-generated works protection under copyright law, most countries continue to strictly interpret the requirement for human authorship, as seen in the EU, US, UK, and Australia. These varied approaches highlight the legal complexities of granting copyright to AI-created works, and highlight the necessity for a balanced legal response that acknowledges AI's potential for innovation while respecting traditional concepts of authorship. Future developments in AI technology will likely prompt further legislative and judicial scrutiny, as policy-makers and courts grapple with the implications of AI's creative and inventive capabilities within intellectual property law. For this study, the author reinforces the need for a harmonised framework that can adapt to the rapid advancements in AI and provide clear, equitable guidance on authorship rights.