Copyright at the Edge of Creativity: Authorship, Infringement, and Royalties in the Age of AI Music — A Nigerian Law Perspective

DOWNLOAD

Introduction

Artificial intelligence is rapidly transforming how music is created and reproduced, outpacing the legal systems designed to protect human creativity. Although computer-assisted music production is not new and has existed since the 1950s, recent advances in generative AI have enabled systems to produce songs, vocal simulations, and musical arrangements with minimal human input. These developments have revived longstanding copyright questions concerning authorship, originality, ownership, and royalty allocation.

Recent developments across the global music industry illustrate the practical and commercial stakes. AI-generated versions of songs such as Stromae's ‘Papaoutai’, Fave's ‘Intentions’, and the controversy surrounding the ‘I Run’ track associated with Jorja Smith raise a number of related legal questions: can AI-generated works qualify as original works? Where music is created with AI as a tool, when does the human contribution suffice for authorship? Can works generated entirely by AI attract copyright protection at all? Does creative transformation excuse the unauthorized use of existing music? And where an AI-generated version achieves commercial success, who is entitled to the resulting royalties? Beyond these specific questions, how can the law protect and reward human creativity without constraining technological innovation?

These questions arise across multiple legal systems and have attracted a range of foreign judicial decisions and industry responses. This article approaches them from the perspective of the Nigerian Copyright Act, 2022 (the "Act"), which modernized Nigeria's copyright framework for the digital environment but does not expressly regulate AI-generated works. As a result, issues relating to authorship, infringement, derivative use, and royalty entitlement must largely be resolved through statutory interpretation and established copyright principles.

Against that background, this article examines how Nigerian copyright law applies to AI-generated music, focusing on the subsistence of copyright, authorship, infringement, and royalties under the Act, together with the emerging legal and practical challenges the technology presents.

The Purpose of Copyright Law

The fundamental rationale of copyright law is to reward creative labour, incentivize innovation, and ensure economic returns flow to those who invest time, skill, and resources in creating original works. The Act explicitly affirms this purpose, identifying as one of its core objectives the protection of authors "to ensure just rewards and recognition for their intellectual efforts," and the strengthening of enforcement in the digital environment.

Artificial intelligence disrupts this balance by generating outputs that appear creative yet do not fit neatly within traditional notions of authorship, compelling copyright law to confront questions for which it was not originally designed.

Works Eligible for Copyright under the Act

Any copyright analysis under the Act begins with whether the subject matter qualifies for protection. Section 2 provides that copyright may subsist in musical works and sound recordings, which together constitute the core copyrights embodied in a song. These categories remain legally distinct, such that different rights may be implicated depending on whether the underlying composition or a particular recorded performance is in issue.

However, eligibility is not automatic. Section 2(2) of the Act imposes two cumulative conditions. First, “some effort” must have been expended to give the work an original character. Secondly, the work must be fixed in a medium of expression from which it can be perceived, reproduced, or communicated, whether directly or with the aid of a machine. This wording represents a shift from the “sufficient effort” standard used under the previous Act, signaling a lower but still meaningful threshold for originality. The Act itself does not define what constitutes “some effort,” leaving its precise content open to interpretation.

This ambiguity is not unique to Nigerian law. In Allen v. Perlmutter, the United States Copyright Office refused to register an AI-generated image despite the applicant's use of over 600 prompts. The Board concluded that the human input did not amount to authorship of the AI-generated elements of the work. Further developments are anticipated, and the boundary between qualifying human effort and mere use of technology therefore remains unsettled. The United States Copyright Act does not mention AI, and the requirement of human authorship has been developed through judicial interpretation of the statute. The U.S. Copyright Office has issued a series of reports and guidance examining the implications of artificial intelligence for copyright. Although Congress has considered related proposals, no comprehensive legislative framework has yet been enacted. Similarly, Germany’s copyright statute protects only works that constitute the author’s own intellectual creation, a standard expressly stated in the legislation and further developed through judicial interpretation. At the EU level, the AI Act (2024) imposes obligations on AI developers but does not amend copyright law. The United Kingdom provides one of the few statutory provisions addressing computer-generated works in Section 9(3) of the Copyright, Designs and Patents Act 1988. However, the Government has proposed that this provision be repealed on the basis that it is no longer adequate for the generative AI era. Nigeria’s reliance on general principles is therefore not a deficiency peculiar to the Act; it reflects the position adopted across most major copyright systems.

While the Act accommodates technological aids in fixation and reproduction, the requirement of "effort" retains a human dimension. The statute does not recognize ideas, processes, or systems, including technological processes, as copyrightable subject matter in themselves. As a result, where music is generated entirely by artificial intelligence without meaningful human creative input, the relevant question is not whether the output fits within a recognized category of works, but whether the human effort necessary for copyright to subsist can be identified at all. On that basis, fully AI‑generated musical works, in the absence of human creative contribution, are unlikely to qualify for copyright protection under the Nigerian Copyright Act.

Can AI Be an Author?

Section 28 provides that, subject to agreement, copyright initially vests in the author of a work, while Section 5 ties subsistence of copyright to the nationality or habitual residence of the author. These provisions presuppose that an author is a legal person capable of holding and enforcing rights.

Artificial intelligence systems do not meet this requirement. They lack legal personality under Nigerian law and cannot own property, hold rights, or receive royalties. While the Act recognizes both natural persons and corporate bodies as possible authors, it does not extend that recognition to artificial intelligence systems. Accordingly, AI cannot, in itself, be regarded as an author for the purposes of copyright law in Nigeria.

This position finds support in comparative jurisprudence. In Thaler v. Perlmutter, the United States Court of Appeals for the District of Columbia Circuit held that copyright protection is limited to works authored by human beings, and that a machine cannot be recognized as an author under the law. Notably, the claimant had presented the work as having been generated autonomously by an artificial intelligence system, without any human authorship, and had listed the AI as the sole author of the work. The same principle had been tested and rejected in a different context in the earlier case of Naruto v. Slater. There, the United States Court of Appeals for the Ninth Circuit held that a crested macaque monkey that had autonomously operated a photographer's camera and taken its own photographs had no standing to bring a copyright infringement claim, because the Copyright Act does not extend authorship rights to non-human entities. Just as the monkey's act of pressing the shutter did not constitute legally cognizable authorship, the generative output of an AI system does not satisfy the human authorship requirement, regardless of how sophisticated or autonomous that output may appear.

Human-AI Collaboration in Music

Artificial intelligence is now widely used by artists as a creative tool in music production. A good example is pop artist Lauv, who worked with AI startup Hooky to produce a Korean version of Love U Like That by using AI voice modelling to transpose a newly recorded translation into his own vocal style. Another striking example is that of country music legend Randy Travis, who suffered a stroke in 2013 that left him largely unable to sing or speak, and nevertheless released his first new song in over a decade in May 2024. The track Where That Came From, was written by professional songwriters and initially recorded using a guide vocal by another singer. This recording was then transformed through an AI model trained on Travis’s past performances to recreate his distinctive voice. Both examples illustrate how AI can support and extend human creative expression.

Section 108 of the Act provides useful guidance by defining the author of a sound recording as the person by whom the arrangements for its making were undertaken. The emphasis on “arrangements” shifts the focus from mere execution to who directed how the work was created. In the context of AI music, this allows recognition of the human as the relevant creator where that person shapes inputs, selects outputs, or modifies the resulting composition. In such cases, it may be prudent for creators to retain evidence of their contribution to the creative process, including records of prompts, repeated revisions, and editorial decisions.

Where a human creator exercises judgment over the form, structure, or expression of a work, the use of AI as a tool does not displace that role. This position, however, does not eliminate all legal risk. Even where sufficient human creative control is established, outputs may still be influenced by the material on which such systems are trained. This introduces a layer of uncertainty, particularly where the resulting work reflects pre‑existing musical elements not directly created by the user, a concern already reflected in recent litigation brought by major music companies against AI music platforms.

AI and Existing Music

A third category of AI-generated music, in addition to fully AI-generated works and human-AI collaboration, emerges where artificial intelligence is used not to produce entirely new works, but to recreate, remix, or modify existing songs. In such cases, the resulting output typically draws on identifiable elements of a prior work, such as its melody or compositional structure.

This is illustrated by the renewed circulation of Papaoutai, written and performed by Belgian artist Stromae and released in May 2013 as the lead single from his album Racine carrée. In early 2026, an Afro‑soul version of Papaoutai went viral online, reportedly produced using AI tools. This version retained recognizable elements of the original work.

The Act addresses such situations through its concept of adaptation. Section 108 defines an adaptation as the modification of a pre‑existing work or its alteration for different forms of exploitation. The right to make adaptations rests exclusively with the copyright owner. Section 9 confers on the owner of a musical work the right to authorize adaptations, while Section 12 protects the reuse or reproduction of sound recordings. Where AI is used to generate a version of an existing song, the need for authorization arises irrespective of the technology employed.

AI-generated versions of well-known songs have circulated widely online, sometimes without prior authorization. In other instances, artists have moved to formalize such versions through collaboration, as Nigerian artist Fave did with her song Intentions. The legal position, however, remains unchanged regardless of outcome: altering an existing work, whether by human effort or through AI, does not displace the requirement for consent. Where a generated work incorporates recognizable elements of a protected song, permission from the rights holder remains the foundation of lawful use.

Infringement and Fair Dealing in AI‑Generated Music

Under Section 36 of the Act, infringement occurs where a person, without the consent of the copyright owner, carries out any act reserved exclusively to that owner. It is not necessary for an entire work to be copied. Infringement arises where a substantial part of a protected work is taken, whether directly or in a form recognizably derived from it. The assessment focuses on whether the portion used captures elements central to the originality of the work.

Where a generated output incorporates recognizable features of an existing song without authorization, the use falls within the scope of infringement. In GEMA v OpenAI, the Munich Regional Court held that where AI systems reproduce recognizable song lyrics in response to user prompts, it amounts to unauthorized reproduction and communication to the public under German law. The case makes clear that it is sufficient for the protected content to remain identifiable, even if it is embedded in the system in a technical or indirect form.

However, not every use of a copyrighted work is infringing. Section 20 of the Act provides for fair dealing, which permits limited use of protected works for purposes such as research, private use, criticism, review, and parody. Whether a particular use is fair depends on factors including the purpose and character of its usage, the nature of the work, the amount and substantiality of the portion used in relation to the work as a whole, and the effect of the use upon the potential market or value of the work.

Fair dealing is unlikely to offer broad protection for most AI-generated versions of existing music. Where substantial elements of a song are used and the resulting output is commercially distributed, the use will typically fall outside its scope.

Royalties and Rights Holders

The entitlement to royalties follows from copyright ownership. Where copyright validly subsists, the author or rights holders are entitled to control the exploitation of the work and to receive remuneration arising from its use. Conversely, where no copyright exists, no lawful claim to royalties can arise.

The legal position is relatively clear. Where a work qualifies for copyright protection, royalties accrue to the human creator or rights holder. Where AI is used to generate versions of existing songs, the entitlement remains with the original rights holder. The use of protected elements, whether in the composition or the recording, requires authorization, and any resulting economic benefit is tied to those underlying rights. Artificial intelligence, lacking legal personality, cannot in itself be a recipient of royalties under the Act.

Emerging Legal and Ethical Challenges

The use of artificial intelligence in music production has given rise to a range of legal and ethical concerns that extend beyond traditional questions of copyright subsistence and infringement. Among the most immediate concerns is the capacity of AI systems to replicate the distinctive voice and stylistic features of existing artists. Even where no direct copying of a musical work can be established, the imitation or cloning of a performer’s voice raises serious questions about consent, artistic identity, and the limits of permissible use.

High‑profile disputes have emerged between major record labels, such as Universal Music Group, Sony Music, and Warner Records, and AI music platforms such as Suno and Udio, centered on allegations that these systems were trained on copyrighted works without authorization. The GEMA v. OpenAI judgment, discussed above, reinforces these concerns. The court's finding that AI models can memorize and reproduce protected works has direct implications for how training practices are assessed.

Beyond disputes over infringement, AI has also exposed vulnerabilities in the enforcement of copyright itself. Folk musician Murphy Campbell discovered in January 2026 that AI-generated covers of her songs had been uploaded to her Spotify profile without her consent. In a separate incident, a third party exploited gaps in audio content recognition systems to file copyright claims against her original YouTube recordings, which were subsequently withdrawn and the responsible user banned. The episode illustrates how existing enforcement mechanisms can be manipulated to appropriate both creative output and the economic benefits attached to it, even where the underlying works belong entirely to the targeted artist.

Artists have also begun to respond to these challenges in different ways. In some cases, AI‑generated content has been resisted through takedowns and public objections. For instance, the controversy surrounding Jorja Smith came from the viral AI‑generated track “I Run” by the producer HAVEN., which featured vocals widely perceived as an imitation of her voice. Following takedown actions by her label and industry bodies, the song was removed from streaming platforms and later re‑released with a different vocalist.

Many of these issues can be traced to the data used to train AI systems. Where large datasets are compiled from existing musical works, often without clear licensing frameworks, the boundary between independent creation and indirect appropriation becomes increasingly difficult to define. This creates uncertainty for artists, rights holders, developers, and platforms alike. The uncertainty is compounded by the fact that the Act does not directly address AI‑generated music or the use of copyrighted works in training such systems.

Conclusion

The questions examined in this article find their answers, at least in part, within the existing framework of the Act. Copyright protection remains grounded in human creative effort; the reuse or alteration of existing works requires authorization; and economic benefits flow to recognized rights holders. Artificial intelligence does not yet displace these foundational rules.

What the Act does not resolve are the more difficult questions at the margins: the sufficiency of human input in AI‑assisted works, the use of unlicensed material in training, and the effectiveness of existing enforcement mechanisms in addressing deliberate misuse. These gaps are unlikely to be resolved through statutory interpretation alone. Whether they are addressed through judicial interpretation or legislative reform, the central challenge will be preserving the legal recognition and economic value of human authorship as technology-enabled creativity continues to expand and become more integrated into the creative process. How Nigeria approaches that reform may inevitably be shaped by developments in jurisdictions grappling with the same questions. Ultimately, the responsibility remains its own. Disclaimer: The article provides general information only and should not be construed as legal advice. References

1. Nigerian Copyright Act 2022

2. Copyright, Designs and Patents Act 1988 (UK), ss 9(3)

3. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 (EU AI Act), entered into force 1 August 2024.

4. United States Copyright Office, Copyright and Artificial Intelligence.

5. Allen v. Perlmutter, Case No. 1:2024cv02665 (United States District Court for the District of Colorado, filed 2024)

6. Thaler v. Perlmutter, No. 23-5233 (United States Court of Appeals for the District of Columbia Circuit, 18 March 2025); certiorari denied, Supreme Court of the United States, No. 25-449 (2 March 2026).

7. GEMA v. OpenAI, Case No. 42 O 14139/24 (Regional Court of Munich I, 11 November 2025) — judgment not yet final.

8. Naruto v. Slater, 888 F.3d 418 (9th Cir. 2018)

9. BBC, ‘Warner Music Group settles lawsuit with AI music firm and launches joint venture’

10. All About Law, ‘Suno Lawsuit Exposed: Warner Secures Multi‑Million Dollar Settlement Over “Stream‑Ripping” Claims’

11. Music Business Worldwide, ‘A folk musician had her voice cloned by AI — and her recordings claimed by a copyright troll’

12. The FADER, ‘HAVEN. vs. Jorja Smith: How “I Run” will shape AI music’s future’

13. Stromae, ‘Papaoutai’, Racine carrée (Universal Music, 2013)

14. CBS News, 'More Than a Decade After a Stroke, Randy Travis Sings Again, Courtesy of AI' (May 2024)

15. Billboard, ‘How Lauv Used AI to Translate His New Single into Korean’

16. NotJustOk, ‘Fave Explains Why She Made the AI Version of Urban Chords’ “Intentions” Official’ (25 November 2025)

17. Herbert Smith Freehills, 'UK Government Report on Copyright and AI Concludes More Evidence is Needed Although s9(3) CDPA Could Go' (March 2026)