Open Access Research Article
White Black Legal – International Law Journal · ISSN 2581-8503
AI TRAINING ON COPYRIGHTED MUSICAL WORKS: LESSONS FOR CAMEROON FROM INDIA’S COPYRIGHT FRAMEWORK
Read the Full Research Paper
Access the complete open-access article in PDF format. No login is required.
Abstract
AI TRAINING
ON COPYRIGHTED MUSICAL WORKS: LESSONS FOR CAMEROON FROM INDIA’S COPYRIGHT
FRAMEWORK
AUTHORED BY - KINANG SONG RICK MARCEL[1]
1. Introduction
Artificial intelligence
has moved from a supporting tool in music production to a technology capable of
composing, arranging and reproducing musical material with minimal human
in-volvement.[2] This shift places direct pressure on a principle that has
anchored copyright law since the nineteenth century: that protection attaches
automatically to an original work, with-out
formality, and cannot
lawfully be exploited without the author’s
consent.[3] The international framework built around this principle – the Berne
Convention, later reinforced by the Agree-ment on Trade-Related Aspects of
Intellectual Property Rights (TRIPS) and the WIPO Copy-right Treaty (WCT) –
extended automatic protection into the digital environment, but none of these instruments anticipated a technology capable of learning
from, and reproducing patterns drawn from, vast quantities of copyrighted material at
industrial scale.[4]
The result is a legal
gap with real economic weight. AI
developers routinely assemble training datasets that include copyrighted songs
and recordings, without a settled
answer as to whether such use requires a licence, falls
within an existing
exception, or escapes
copyright altogether. For developing music markets, unlicensed training risks displacing the very revenue streams – licensing, royalties,
collective remuneration – that a national music industry needs to grow, at
precisely the moment those markets are beginning to gain international
visibility through digital streaming.
The stakes are not
evenly distributed. Large technology
companies and well-resourced entertainment businesses tend to possess disproportionate access to computing
power, training data and legal expertise, positioning them to capture
much of the commercial value AI-generated music creates; independent musicians, particularly in developing markets, may find themselves competing against AI-generated content produced at a cost and speed
no individual artist can match, without
a corresponding share
of the resulting economic benefit. At the same time, AI can lower the barriers
that have historically excluded musicians without
access to ex-pensive studios or professional
production infrastructure – a genuine democratising potential that any
regulatory response must be careful not to foreclose. A sound legal framework must hold both concerns simultaneously: protecting creators
against uncompensated large-scale ex-ploitation of their work, while preserving the opportunities
AI tools offer to musicians who have historically had the least access to
professional production infrastructure.
This article compares
how India and Cameroon confront this specific problem: the use of copyrighted musical
works to train
AI systems. It argues
that neither jurisdiction currently has a
framework adapted to this use, but that the disparity between them is
instructive. India’s considerably more developed music industry and copyright infrastructure have produced insti-tutional and judicial responses –
however incomplete – that Cameroon, still building its own framework under Law
No. 2000/011 and the regional OAPI
system, can draw on. The article does not suggest that India’s model be transplanted wholesale; it identifies specific institutional
capacities that offer
a workable template
for reform in Cameroon.
Section 2 sets out the shared
international framework within which both countries operate. Sections 3 and 4 examine, re-spectively,
Indian and Cameroonian copyright law as each applies to AI training. Section 5 draws out what, specifically,
Cameroon can learn from India’s experience, and Section 6 sets out concrete
recommendations.
2.
The International Framework
Both India and Cameroon
operate within the same international copyright architecture. India is a party to Berne, TRIPS, the WCT
and the WIPO Performances and Phonograms Treaty (WPPT) in its own right;
Cameroon is bound by the same instruments and additionally par-ticipates in the regional
system administered by the African
Intellectual Property Organization (OAPI).[5]
This shared international foundation means that both countries guarantee
automatic protection to musical works and sound recordings, and both are bound
by TRIPS’ minimum enforcement standards, including
the availability of civil remedies,
injunctions and damages for infringement.[6]
Article 9(2) of TRIPS
confirms that copyright protects expression rather
than the ideas,
pro-cedures or methods underlying it – the idea-expression dichotomy that underlies every domestic
infringement analysis this article later examines.[7] The WCT, for its part, extended Berne-style protection into the digital
environment, addressing computer
programs, databases and the right of
communication to the public that online streaming engages.[8] What none of these instruments does – because none was drafted
with the possibility in mind – is specify
whether the automated ingestion of copyrighted works
into an AI training dataset
constitutes an act of reproduction re-quiring authorisation, or whether it might instead fall
within some analogue of the exceptions Article
9(2) and the Berne three-step test otherwise permit.[9] That silence is not an oversight so much
as a structural limitation:
international copyright instruments are renegotiated rarely and
slowly, while the technology this article examines
has developed within
a handful of years. The practical consequence is that the
question is left entirely to domestic law – which is precisely where India and
Cameroon diverge most sharply, as Sections 3 and 4 show.
3.
India: A Head Start Without
Settled Answers
India’s copyright
system rests on the Copyright
Act 1957, amended
several times, most significantly in 2012 to implement the WIPO Internet
Treaties. Section 2(d)(vi) of the Act provides
that, for a computer-generated literary, dramatic, musical or artistic work,
the author is “the person who causes the work to be created.”[10] India is therefore among a small number of
jurisdictions whose statute expressly contemplates computer-generated works – a
provision Cameroon’s legislation simply does not have.
The difficulty is that this provision predates
gen-erative AI by decades:
it was drafted for software
that followed instructions a human had sub-stantially
specified, not for systems that produce a complete composition from a short
prompt while independently determining melody, harmony and arrangement. Whether a person who supplies such a
prompt “causes the work to be created” within the meaning the provision
in-tended remains untested, and no Indian court has resolved the question in
relation to music.
The Act’s fair-dealing exception, Section 52, permits certain uses – private
use, research, criticism and review
– without infringing copyright.[11] Whether
this exception extends
to large-scale commercial AI
training is genuinely uncertain: it was drafted around individual, largely
non-commercial uses, not the automated ingestion of millions of recordings by a
commercial developer, and no Indian court has ruled directly on the question.
Ownership adds a
further layer of difficulty. Section
18 permits an author to assign copy-right, and
Section 19 sets out the formal requirements for a valid assignment, including that it be in writing, signed, and specify the rights, duration and
territorial extent assigned.[12] This assignment framework is what would allow an AI developer to obtain rights to musical
works by agreement rather than relying on an exception – but where a
work’s authorship is itself unsettled, as Section 3.1 explains, the assignment
framework built on top of that authorship becomes correspondingly harder to
apply with confidence, since a licensee cannot be certain what, precisely, it
is acquiring, or from whom.
Performers receive
separate protection under Section 38, of particular relevance to AI voice
cloning: because a performer’s right
exists independently of the copyright in the underlying composition, an AI
system that reproduces a performer’s recognisable vocal characteristics may engage Section 38 even where it does not copy any specific
existing recording.[13] Section 57 separately protects an author’s
moral rights of attribution and integrity, engaged whenever AI-generated content
is falsely attributed to a musician
or distorts an existing work in a manner
prejudicial to their reputation.[14] Neither provision was drafted with AI in
mind, and India has no legislation addressing unauthorised
voice cloning or deepfake music as such, leaving affected performers to construct a claim from whatever combination of these general
provisions happens to fit the facts.
Where India’s
framework does offer something concrete
is institutional. Sections
33 to 36 regulate copyright societies – India’s collective management
organisations – requiring registration with the Registrar of Copyrights and
imposing obligations of transparent royalty dis-tribution among members.[15] These societies administer performance and
broadcasting rights at scale on behalf of composers, lyricists and publishers,
and this existing infrastructure of membership
records, repertoire databases and royalty-distribution mechanisms is precisely the institutional capacity that any collective licensing
solution to the AI-training problem would require. Whether India’s copyright societies could realistically be
adapted to license AI train-ing use, rather than only performance and
broadcasting, remains an open question – but the statutory framework permitting
collective administration is, unlike an AI-training exception, already in
place.
India’s judiciary
has also begun engaging with digital-era exploitation more broadly, even
without addressing AI training directly. In Eastern Book Company v D.B. Modak, the Supreme Court articulated a “skill and judgment” threshold for originality,
rejecting both a bare “sweat of the brow” standard and a strict
originality requirement, and adopting an intermediate test relevant to how AI-processed or AI-assisted works
might later be assessed for originality.[16] In R.G. Anand v Delux Films, the Supreme Court articulated the idea-expression dichotomy
in the Indian context, holding
that copyright protects the particular expression of an idea rather than the
idea itself – a distinction that will become central once Indian courts are
asked whether an AI system’s output
merely reproduces an unprotectable style or genre, or the protectable
expression of a specific work.[17] In Super
Cassettes Industries Ltd v MySpace Inc, the Delhi High Court at first instance addressed intermediary liability for infringing content hosted on a digital platform, a body of reasoning with direct relevance to how a court might
later approach a platform hosting AI-generated music
– although the Division Bench
subsequently reversed that finding on appeal, holding
that My Space qualified for safe-harbour protection as an intermediary, a reminder that Indian jurisprudence on platform liability
is itself still unsettled.[18] None of
these cases resolves the AI-training question directly, but together they show a judiciary already
accustomed to reasoning about originality and digital exploitation – a
comparative advantage Cameroon’s courts have not yet had the same opportunity
to develop, given the absence of a comparable body of jurisprudence.
Taken together, India’s
strengths are real: comprehensive statutory protection for musical works and
sound recordings, express recognition of performers’ and moral rights, an established collective
management infrastructure,
and a judiciary
with some experience of digital-era
reasoning. Its limitations are equally real: Section 2(d)(vi) does not resolve how much human involvement its “causes the work to be created”
test requires; no provision
addresses AI train-ing
specifically; and no legislation addresses
voice cloning or deepfake music as such. India’s statutory head start on computer-generated works has not, in practice,
translated into settled answers
for the AI-specific questions this article examines – but it has produced
institutions and judicial experience that Cameroon does not yet have.
4.
Cameroon: A Sound Principle, an Absent Institution
Copyright in Cameroon
is governed principally by Law No. 2000/011
of 19 December 2000 on Copyright and Neighbouring Rights,
which protects “musical
compositions with or without
lyrics”[19]
and grants authors the exclusive right to exploit their work in any form and to
derive pecuniary profit from it[20] – a formulation broad enough, on its face,
to extend to AI-related exploitation despite
predating both digital
streaming and generative AI by well over a decade.
The Law recognises both economic
and moral interests
in broadly the same structure
as the international framework examined in Section 2. Economic rights allow authors to control
reproduction, representation, adaptation and distribution of their work; moral rights
protect the personal relationship between author and work, engaged
in essentially the same circumstances as under Indian law – false
attribution of AI-generated content to a musician, or unauthorised digital
manipulation of an existing work in a manner harmful to the author’s
reputation. The Law also recognises neighbouring rights for performers, producers of phonograms and broad-casting organisations, the domestic counterpart of the Rome Convention and WPPT framework:
performers’ contributions are recognised as distinct from the underlying
composition, mean-ing voice-cloning technology raises the same analytical difficulty for Cameroonian law that it
raises for Indian law under Section
38.
The 2000 Law contains
no provision addressing AI-generated works, computer-generated authorship, or
AI training specifically. Where a
Cameroonian musical work is copied into an AI training dataset, the general
reproduction right is engaged in principle, but whether any existing exception
could apply to such use is untested,
and considerably more uncertain than in
India given the absence of a comparable body of jurisprudence. Cameroon’s courts have not yet had occasion to address digital
exploitation at the scale India’s have. A
further complication, largely absent from the Indian context, concerns
Cameroon’s musical heritage: genres such as Makossa and
Bikutsi are culturally and commercially significant, and the 2000 Law – built,
like all copyright statutes, around individual authorship – was not drafted to
answer how AI processing of musical patterns and traditions developed
collectively over generations, rather than owned by any single identifiable rights holder, should
be treated.
The sharper
problem, however, is institutional rather
than statutory. Effective enforcement requires digital monitoring
capacity, reliable rights databases, metadata systems and accessible judicial
remedies – infrastructure that remains considerably less developed in Cameroon
than in India. Collective management,
which the 2000 Law contemplates, has not developed the scale or technological sophistication of India’s
copyright societies, even though individual Cameroonian creators
frequently lack the practical means to monitor every use of their own work
without such an institution acting on their behalf. Cameroon’s participation in OAPI situates its national
legislation within a regional framework shared with other member states, which
offers a channel for coordinated reform but does not, on its own, supply the
domestic monitoring and licensing capacity Cameroon currently lacks. This institutional gap is decisive: even a well-drafted AI-training provision would have little practical effect without an institution
capable of administering the licences, royalty flows and monitoring it
would require.
5.
What Cameroon Can Learn From India
The comparison yields a
specific rather than general lesson. Cameroon
does not lack sound statutory principle: Section
15(1) of the 2000 Law states the exclusive right of exploitation in terms at least as protective as India’s
Copyright Act. What Cameroon lacks is
the institu-tional machinery to make that principle operative against a
technology the 2000 Law never anticipated.
Three elements of
India’s experience are directly transferable.
First, India’s copyright so-cieties demonstrate that collective
management infrastructure – membership registries, reper-toire databases,
transparent distribution rules – can be built at national scale and later
adapted to new forms of exploitation; Cameroon’s collective management system,
still comparatively nascent, could be developed along the same institutional
lines, drawing on India’s regulatory model under Sections 33 to 36 rather than
starting from first principles. Second,
India’s Sec-tion 2(d)(vi), whatever its interpretive difficulties, shows that a
legislature can at least name computer-generated authorship as a distinct
statutory category; Cameroon’s legislature could adopt a comparable provision – ideally drafted
with more precision than India’s, expressly
ad-dressing the degree
of human contribution required – rather
than leaving the question to silence.
Third, India’s judiciary’s willingness to engage with platform liability
and digital exploitation, however incomplete its answers to AI-specific questions
remain, offers Cameroonian courts
a body of comparative reasoning to draw on as similar
disputes eventually reach them.
None of this implies
that India’s framework is adequate on its own terms – Section 4 of this article
and the underlying dissertation both show that it is not. The claim is narrower: between two
jurisdictions that share the same international obligations but differ sharply
in institutional maturity, the more developed system’s partial answers are a
more useful starting point for reform than a blank statutory page.
6.
Recommendations and Conclusion
Four recommendations
follow directly from this comparison. First,
Cameroon’s legislature should amend Law No.
2000/011 to include an express
provision addressing AI-generated and AI-assisted works,
specifying the degree of human creative
contribution required for authorship
to attach – learning from, rather than replicating, the interpretive
uncertainty Section 2(d)(vi) has produced in India. A graduated test, distinguishing assistive use of AI from
substantially autonomous generation, would give Cameroonian courts a clearer
starting point than India’s courts currently have.
Second, Cameroon should
prioritise building collective management capacity specifically capable of administering AI-training licences, modelled on the institutional structure of India’s copyright societies under
Sections 33 to 36 of its Copyright Act rather than reproducing the whole of Indian law. A licensing scheme of this kind – potentially built
on a compulsory or extended collective licence, allowing AI
developers to obtain authorisation to train on a defined repertoire in exchange
for a set remuneration flowing
back to rights
holders – offers
a more administrable path than requiring
individual authorisation from every affected
musician, particularly for a market where individual creators frequently
lack the resources to negotiate directly with AI developers.
Third, performer
protection deserves separate legislative attention in both jurisdictions, given
that voice cloning implicates personality and reputational interests distinct
from the un-derlying musical composition; a dedicated statutory remedy for unauthorised
voice cloning, rather than one assembled indirectly from performers’ rights
and moral rights
provisions never drafted
with this technology in mind, would benefit both India and Cameroon, though Cameroon
in particular given its less developed jurisprudence in this area.
Fourth, both countries would
benefit from pursuing
this reform through
regional and inter-national cooperation – Cameroon
through OAPI, India bilaterally and through existing
WIPO channels – given that AI
training and music distribution both operate across borders that a purely
domestic reform cannot fully address. A
reform agenda pursued in isolation by either country risks being outpaced by an
AI industry that operates, by its nature, across all of them at once.
Artificial intelligence
has not created a new copyright principle; it has exposed how much the enforcement of an old one depends
on institutional capacity
rather than statutory text alone. India’s experience shows that even a legal head start does not resolve every question AI raises.
Cameroon’s experience shows that sound statutory principle, without the
institutional means to enforce it, offers
creators only nominal protection. The
comparison between them sug-gests that the most useful reform for Cameroon lies
not in copying India’s law wholesale, but in building the specific
institutional capacities – collective management, statutory clarity on
computer-generated authorship, and judicial familiarity with digital
exploitation – that have allowed India’s imperfect framework to function in
practice.
[1] Kinang Song Rick
Marcel, LLM (One Year Course), School of Law, CT University, Punjab, India. This article draws on the author’s LLM
dissertation, Artificial Intelligence,
Globalisation and the Music Industry: A
Comparative Legal Study of Copyright Protection in India and Cameroon,
undertaken under the supervision of Dr. Cheena Abrol,
Assistant Professor, School of Law, CT University.
[2] See generally
World Intellectual Property
Organization, WIPO Technology Trends 2019: Artificial Intelligence (WIPO 2019).
[3] Berne Convention for the
Protection of Literary and Artistic Works (adopted 9 September 1886, as
revised) art 5(2).
[4] Agreement
on Trade-Related Aspects
of Intellectual Property Rights (TRIPS) (adopted 15 April
1994, entered into force
1 January 1995) art 9; WIPO Copyright Treaty (adopted 20 December 1996,
entered into force
6 March 2002).
[5] See Organisation Africaine de la Proprie´te´ Intellectuelle, Bangui Agreement (as revised 1999).
[6] TRIPS (n 3) arts 41–61.
[7] TRIPS (n 3) art
9(2).
[8] WCT (n 3) arts 4, 5, 8.
[9] Berne Convention (n 2) art 9(2);
TRIPS (n 3) art 13.
[10] Copyright Act, 1957, §2(d)(vi).
[11] Copyright Act, 1957, §52.
[12] Copyright Act, 1957, §§18–19.
[13] Copyright Act, 1957, §38.
[14] Copyright Act, 1957, §57.
[15] Copyright Act, 1957, §§33–36.
[16] Eastern Book Co. v. D.B. Modak, (2008)
1 S.C.C. 1 (India).
[17] R.G. Anand v. Delux Films, A.I.R. 1978 S.C. 1613 (India).
[18] Super Cassettes Indus. Ltd. v. MySpace Inc.,
2011 (48) P.T.C. 49 (Del.) (India), rev’d, MySpace Inc. v. Super Cassettes Indus. Ltd.,
2016 SCC OnLine Del 6382 (India).
[19] Law No 2000/011
of 19 December 2000 Relating to Copyright and Neighbouring Rights (Cameroon) s 3(1)(b).
[20] Law No. 2000/011, §15(1).
How to Cite This Article
KINANG SONG RICK MARCEL, AI TRAINING ON COPYRIGHTED MUSICAL WORKS: LESSONS FOR CAMEROON FROM INDIA’S COPYRIGHT FRAMEWORK., White Black Legal – International Law Journal, ISSN: 2581-8503, Vol. Volume 4, Issue VOLUME 4 ISSUE 2, October 2026, pp. 16-27. Available at: https://www.whiteblacklegal.co.in/public/details/ai-training-on-copyrighted-musical-works-lessons-for-cameroon-from-indias-copyright-framework
Author & Publication Record
Authors: KINANG SONG RICK MARCEL
Registration ID: 107125 | Published Paper ID: WBL7125
Year: Oct- 2026 | Volume: 4 | Issue: 2
Approved ISSN: 2581-8503 | Country: Delhi, India
Page No.: 16-27
Full Text Preview
Open in New Tab
Copied