Authorship without Authors? AI-Generated Music Copyright in Singapore and the UK

Writers: Hannah Lau and Chiara Ng

Editors: Alyssa Tan and Sijia Cai

[2025/26 Editorial Committee]

Introduction

The opening of any social media platform is now accompanied by an onslaught of generative artificial intelligence (AI) content. With this rise in popularity, there has been an influx of creative works that are generated from AI systems that train on large-scale datasets of existing academic and creative work.^[1] Under copyright law, questions have arisen regarding whether AI-generated music can and should be protected under the law. This article will examine (1) who owns generative artificial intelligence music, and (2) whether AI-generated music should be protected at all through the lens of UK and Singapore law.

Who owns Generative Artificial Intelligence Music?

Authorship in Singapore and the UK

Both the UK and Singapore’s intellectual property laws have rooted copyright in human authorship, but differ in their treatment of computer-generated work.

In the UK, the Copyright, Designs and Patents Act 1988 (CDPA) section 9(1) defines ‘author’ as ‘the person who creates [the work].’ In Thaler v Comptroller-General, there is heavy emphasis on human authorship.^[2] Lord Hoffmann reiterated that ‘the inventor must be a natural person’.^[3] As DABUS is not a person, it was not the inventor. While the developer of a generative AI program did create the program, the way they train the program is derived from large data sets, typically taken from other creators. A well-known example is the AI-generated song ‘Heart on My Sleeve’, which featured the voices of Drake and the Weeknd.^[4] 

Singapore shares this strong human authorship requirement. Under the Copyright Act 2021 (CA), the first owner of copyright in a work for sound recording would be the maker of the recording.^[5] Under sections 21 to 23, the maker of a sound recording is the person who owns the first record embodying the sounds when the recording is produced, and it is made when the first record embodying the sounds is produced.^[6] The IPOS guidance echoes this sentiment as well, stressing the importance of human creative input.^[7]

While the UK has a statute to protect computer-generated work, Singapore lacks this statute, having no equivalent to section 9 (3) CDPA. In the UK, CDPA 1988 section 9 (3) specifies that for computer-generated works, the author is defined as ‘the person by whom the arrangements necessary for the creation of the work are undertaken’.^[8] This was originally established to encourage investment in early computing systems, reflecting a longstanding acceptance of computer-generated works.^[9] Currently, AI-generated music could arguably fall under this section.

Who owns AI-generated music?

Having discussed authorship, the difficulty remains in proving whether AI-generated work is substantial enough to warrant copyright.

Under UK law, the concept of “substantial part” in copyright law can be applied. This stems from section 16(3)(a) CDPA, which states that ‘references in this Part to the doing of an act restricted by the copyright in a work are to the doing of it in relation to the work as a whole or a substantial part of it’. This examines the qualitative element of a work rather than the quantitative.^[10] In Designers Guild v Russell Williams,^[11] two cases identified where this concept can be applied. One is when an identifiable part of the whole, but not the whole, is copied, and whether this is an infringement relies on the qualitative importance of the part that has been copied. However, if the music can pass the substantial part test, the owner would most likely be the user who created the prompt.

It is possible to argue that the AI user has joint authorship over the generated music. The case of Kogan v Martin^[12] questioned whether there was joint authorship between the person who wrote the screenplay and the person who contributed ideas and helped with the process. It held that joint ownership of copyright is determined on the facts of each case, and that an author can be a person who provides input on expertise and has intertwined intellectual contributions to the work’s original idea. Joint authorship requires collaboration, the contribution must be more than trivial, and the input must be authorial or creative choices. This could mirror the concept of how the user of the AI program submits prompts, which can be considered an intellectual contribution to the generation of the product. However, the extent of creative contribution could be difficult to measure, especially when the music produced is derived from datasets collected from existing creative work.

Singapore utilises the substantial similarity test, seen in Global Yellow Pages v Promedia Directories Pte Ltd (“Global Yellow Pages”).^[13] The facts of the case surround the production of telephone directories and whether the defendant had infringed its copyright by copying and referencing the listings and classifications in its directories. The Court of Appeals held that there was no infringement, as ‘what must be shown is substantial similarity between those elements and only those elements’.^[14] It was additionally emphasised that to copy unoriginal parts of a copyright work will not, on its own, amount to a substantial part. With this emphasis on originality and the copying of original works, the process of AI-generated music that holds the most potential for originality would ultimately be the prompt writer and user of the program.

Should AI-generated music be protected?

A core legal uncertainty lies not just in who owns AI-generated music, but also in whether such music should be protected by copyright at all. The UK adopts a pragmatic statutory fiction under section 9(3) CDPA, while Singapore maintains a more conservative, human-centric approach under the CA 2021. This substantially affects their approaches to AI-generated music.

Human Authorship

Both the UK and Singapore root their copyright regimes in the notion of human authorship. In Thaler, the UK Supreme Court reaffirmed that intellectual property (IP) law fundamentally presupposes a human creator.^[15]

Nevertheless, originality in AI‑generated outputs may arise not from the machine itself but from the human user's creative inputs and constraints, such as prompts, parameters, or stylistic direction.^[16] Section 9(3) CDPA accommodates this view, attributing authorship to the person ‘by whom the arrangements necessary for the creation of the work are undertaken’. Under this view, generative music produced by systems analogous to ChatGPT could still meet the originality threshold if it reflects the user’s intellectual contribution.

In contrast, Singapore’s Court of Appeal in Global Yellow Pages affirmed that originality requires “skill, judgment, and effort” exercised by a human mind.^[17] The Singaporean approach confines copyright strictly within the domain of human intellectual labour. This remains coherent with the rest of Singapore’s approach of meritocracy, to award and protect pure human effort more readily than indirect human input. 

Law as an Instrument of Economic Policy

The UK’s statutory framework reflects a precautionary approach, utilising legislation as a protective tool rather than a merely regulatory one. Section 9(3) CDPA ensures that ownership is attributed to a human or corporate entity, preventing gaps in protection that might undermine commercial exploitation and investment. Clearly, the UK acknowledges the creative industries’ economic value. This aligns well with its historically protective stance towards the creative industries. If originality can be grounded in the human’s preliminary creative decisions rather than the machine’s autonomy, protection becomes consistent with both doctrinal and economic aims.

Singapore’s stance diverges sharply, shaped by its ambition as a global technology and AI hub. The CA omits any equivalent to section 9(3), which seems to deliberately emphasise the role of copyright as an instrument of innovation policy. Keeping AI-generated music in the public domain promotes experimentation, reduces transaction costs and avoids monopolisation of algorithmically generated outputs. Further, Singapore’s primary targets are AI developers and companies that shape the next generation of technology. Access to vast networks of databases is therefore most crucial. As such, the UK values the creative industry’s skill, while Singapore focuses on economic benefits and efficiency.

The utilitarian approach of Singapore is also evident in the fewer moral rights the CA provides when compared to the CDPA. The UK highly values artistic integrity, seen in extensive case law like Sawkins v Hyperion Records^[18] and Walter v Lane.^[19] This is coupled with robust moral rights covered in the CDPA: the right of attribution,^[20] the right against false attribution,^[21] the right to integrity,^[22] and the right to privacy.^[23] In contrast, Singapore has limited legislation on moral rights, adopting only the right of attribution,^[24] the right against false attribution,^[25] and creating the right not to have an altered copy represented as unaltered.^[26] The absence of the right to integrity (which protects against derogatory treatment) and the right to privacy evidently highlights Singapore’s preferred approach of AI as a tool for technological and economic advancement, rather than focusing on the expression of human creativity within it. Protection of AI-generated music takes the back seat while the utilitarian benefits it provides are exploited.

Difficulties in the realm of protection

Despite these different approaches, both face their respective set of difficulties.

Overprotected or underprotected?

In the UK, the most immediate difficulty is that copyright law is built around human creativity. The UK requires the work to be the author’s ‘own intellectual creation’.^[27] Yet combined with section 9(3), it does provide some leeway for AI assistance. The key distinction lies between AI-assisted music and AI-generated music. This is a delicate boundary to define. If AI-generated music receives too much protection, minimal prompts could attract full copyright, and the threshold for originality becomes trivialised. The ruling in Global Yellow Pages^[28] on originality will gradually hold little value. It also strains the Infopaq requirement of creative choice. This raises further questions about whether separate protections should apply to distinct elements of AI-generated music, such as lyrics and musical composition.

Nevertheless, Getty Images v Stability AI suggests that the UK courts are conscious of overstepping,^[29] emphasising that the wording of prompts must ‘make a real difference’ to the output of the AI for a copyright claim to be made.^[30] This approach may not, however, be willingly adopted by their Singapore counterpart. As discussed above, Singapore seeks the rapid advancement of AI. This suggests that the threshold for a copyright claim in AI will be higher and in favour of AI.

Singapore faces the opposite problem. Without protection, investors in generative music platforms are disincentivised, and users who meaningfully shape AI compositions cannot monetise their contributions. A further consideration is the structural relationship between creativity and innovation. Technology innovation rarely emerges in isolation from creative practice. Historically, creative experimentation has driven technological development, for example, the evolution from MP3 encoding to platform-based distribution models such as Spotify. For a forward-looking economy like Singapore, this may mean inhibiting technological advancements, which seems inconsistent with Singapore’s pro-innovation approach.

Other shared problems

Beyond domestic concerns, the divergence between the UK and Singapore approaches poses significant cross-border challenges. The interconnectedness that AI brings to the world promotes cross-border collaboration, which becomes legally uncertain where jurisdictions adopt drastically different approaches. For instance, the UK courts in Lucasfilm v Ainsworth grappled with applying domestic copyright principles to works created across multiple jurisdictions.^[31] Given that music is characteristically borderless, divergent legislative approaches are further likely to have a greater disruptive impact on the music industry than on other fields. This fragmentation remains one of the most pressing challenges faced in the AI music industry.

Conclusion

Despite Singapore’s roots in English law, the above comparison of the UK and Singapore’s legislation raises interesting differences in the policies and societal values that have shaped the development of their respective laws. The UK’s protective stance offers clear short-term benefits under section 9(3). It safeguards existing creative industries, ensures that investment in AI music platforms is commercially viable, and maintains the UK’s reputation as a jurisdiction hospitable to the creative economy. Whereas Singapore encourages rapid experimentation and lowers barriers to entry for AI music development, highlighting the benefits reaped in the long-term and maintaining its status as a leader in the global AI economy. As generative AI continues to reshape the music industry, both jurisdictions should constantly revisit their frameworks to strike a balance that is legally coherent, economically sound and respectful of modern creativity.

References

^[1] Gabrielle Chou and Nicholas Lang, ''The Sound Shift: How Generative AI is Redefining the Music Industry’s Business Model” (ARTEFACT) <https://www.artefact.com/blog/the-sound-shift-how-generative-ai-is-redefining-the-music-industrys-business-model/ > Accessed 13 February 2026.

^[2] Thaler v Comptroller-General of Patents, Designs and Trade Marks [2023] UKSC 49

^[3] Thaler v Comptroller-General of Patents, Designs and Trade Marks [2023] UKSC 49 [63]

^[4] Pujari, Vinish & Wilson, Bindy ‘Copyright and Authorship in AI-Generated Music’ [2023] Journal of Emerging Technologies and Innovative Research. 10. f351-f354.

^[5] Copyright Act 2021

^[6] ibid.

^[7] ‘Copyright 101 Infopack’ (Intellectual Property Office of Singapore, 24 Feb 2023) Accessed 13 February 2026

^[8] Copyright, Designs and Patents Act 1988

^[9] Parish, J. (2025). Time to Repeal Section 9(3) of the Copyright, Designs and Patents Act 1988: New insights from the lobbying and drafting history behind the infamous UK computer-generated works regime. INTELLECTUAL PROPERTY QUARTERLY. Advance online publication.

^[10] Ladbroke (Football) Ltd. v. William Hill (Football) Ltd. [1964] 1 W.L.R. 273

^[11] Designers Guild v Russell Williams [2000] 1 WLR 2416

^[12] Kogan v Martin [2019] EWCA Civ 1645

^[13] Global Yellow Pages v Promedia Directories Pte Ltd [2010] SGHC 97

^[14] ibid.

^[15] Thaler v Comptroller-General of Patents, Designs and Trade Marks [2023] UKSC 49

^[16] Sogut Atilla, ‘Dealing with AI-generated works: lessons from the CDPA section 9(3)’ (2024) 19(1) Journal of Intellectual Property Law & Practice <https://doi.org/10.1093/jiplp/jpad102> accessed 12 February 2026

^[17]  ibid (n. 12) [18(a)ensures]

^[18] [2005] EWCA Civ 565

^[19] [1900] AC 539

^[20] Copyright, Designs and Patents Act 1988, s 77

^[21] ibid. s 84

^[22] ibid. s 80

^[23] ibid. s 85

^[24] Copyright Act 2021, s 371-377

^[25] ibid. s 378-379

^[26] ibid. s 380

^[27] Case C-5/08 Infopaq International A/S v Danske Dagblades Forening [2009] ECR I-6569

^[28] Global Yellow Pages v Promedia Directories Pte Ltd [2010] SGHC 97

^[29] Getty Images v Stability AI [2025] EWHC 2863 (Ch)

^[30] ibid. [203]

^[31] Lucasfilm v Ainsworth [2011] UKSC 39