Composing the Future: Legal Uncertainties Surrounding AI-Generated Music Replication

Artificial Intelligence has created a transformative revolution in our society as it slowly captures various markets. Consequently, with a revolution come dynamic changes that the existing law does not essentially accommodate. Voice cloning of an artist to create songs in their style is an ever-rising trend that has recently garnered traction in India. Technological advancement is the root of a developing country. When changes like this hope to be adopted in a culturally diverse society like India that respects authenticity in artistic works, multiple question are pertinent to be answered. The key question being: is the existing law capable of balancing the interests of either parties or a speedy review is needed by the legislation at the earliest? The author aims to answer these issues and provide possible solutions like fair royalty and licensing agreements.

Keywords: Copyright Law, Royalty & Licensing, Artificial Intelligence, Voice Cloning, Personality Rights.

Introduction

Artificial Intelligence (hereinafter “A.I.”) is going through a remarkable surge in the Indian landscape but it is only recently that the creative community is taking note of the identifiable trend of incorporating A.I. to ease one’s job. Unfortunately, with it also follows certain issues and implications. While A.I. Software like Mid-Journey have created a ground-breaking revolution with their entry in the art community[1], it is only recently that these tremors have been felt in the music industry. Nonetheless, the creator community received clarity to an extent when the U.S. Copyright Office in 2013 rendered artworks created by software like Mid-Journey as non-copyrightable.[2] One can reasonably assume that it is only imminent that this decision would soon stretch to other areas of creation. While the West has been actively filling legal lacunas correlating to A.I., India’s pace has only picked up acceleration.

Recently, A.R. Rahman (hereinafter “Rahman”), a renowned Indian Musician and two-time Oscar Winner, was in talks when he employed an A.I. software in his music creation process. He used A.I. to render the voices of famous deceased artists Shahul Hameed and Bamba Bakya in a song “Thimiri Yezhuda” for a movie titled Lal Salaam.[3] While this received a positive response from fans who had accepted their fate of never hearing from the musicians again, it also instigated a conversation in the creative community as to how one can draw the fragile line between paying homage to an artist as was claimed by Rahman on one hand, and explicitly exploiting the personality rights of the artist (deceased or alive) as was alleged in the criticism the song received on the other hand.[4] This conundrum and its legal ramifications lay the base for what this article tries to answer in the next segments.

A.I. Voice Cloning: Innovation or Infringement?

While Rahman received a mixed response in India for his song[5], A.I. cloned music has not necessarily seen a warm welcome in the Western Music Industry. UMG, a renowned music producing company handling artists like Drake and The Weeknd was in the news after they took down a song created by an A.I. software replicating the singing styles of their aforementioned clients.[6] This take down was on grounds of copyright infringement and stayed in the news for a while, initiating the debate on A.I. and its position in the industry. However, the recent trigger that brought this discussion to worldwide limelight is the news of 200+ artists including the likes of Zayn Malik, Billie Eilish, Katy Perry, Imagine Dragons,[7] and more renowned names signing an open petition for the ban of A.I. voice cloning to create new renditions and covers.

While one may contest this to be a single occurrence, multiple copyright claims are being filed across the globe by artists and their management companies to secure and protect their personality and moral rights, showcasing a dire need to discuss the legal ramifications of this issue.[8] The most pertinent point while discussing the boundaries of A.I. Music is the extent till which A.I. software companies contest the defence of Fair Dealing.

Process of Creating A.I. Sampled Music

A.I. software like Meta’s Audio-Craft are known for sampling multiple songs in the form of codes.[9] These are royalty-free testing[10] where the software utilises pre-existing copyrighted music to learn the patterns and styles of singing and simply incorporates it into a new output as per the individual’s input. This can be seen across the plethora of covers of popular songs uploaded on YouTube and Spotify[11] that have been covered by various famous artists when in actuality they have not sung a single lyric in that song or are paid their due royalty for voicing the song.

In the landmark case of Eastern Book Company v. D.B. Modak[12], the Supreme Court laid down the Doctrine of Modicum of Creativity as the principle for assessing whether a certain artistic work is copyrightable or not. Justice P.P. Naolekar in this Case qua this Doctrine discussed the necessity of having a minimum amount of creativity in one’s work to render a creation copyrightable, as mere labour is not enough. Another test laid down by Justice S. Murtaza Fazal Ali in R.G. Anand v. Delux Films[13], discusses that to determine whether there exists copyright infringement in the impugned work, one must assess whether an individual of ordinary prudence would be of the opinion that the impugned work is a subsequent work to the original created by the plaintiff. Application of these tests to the creation process of A.I. music displays a very clear picture. One can conclude that infringement can be assessed when an ordinary listener can determine similarities between the songs created by A.I. with the existing works of the singer. This may include similarities in tune, singing style, vocals. It may also include recognition of the voice of the artist that forms a part of their personality right.

Analysis of A.I. Recreations of Artists’ Voice or Style

It is a settled principle that ideas are not copyrightable unless they are presented tangibly. Expression in Copyright Law is mandatory, unlike in Patents Law where idea is a strong base for rights of ownership. The same was established in the Eastern Book Company[14] case. This additionally adds to the dilemma of A.I. recreation covers as a plaintiff cannot claim copyright on an idea alleging that they could have produced the same output had they taken up the project. To understand this dilemma better, one needs to assess exactly what is copyrightable in music. Music in its essence is a culmination of chords and rhythms which are played in multiple permutations and combinations. The differentiating factor in any song is the composition and the singer’s unique vocal identity.

There stands no chord or harmony that has not been utilised till now, the difference only lies in the composition of the same. This can also be seen in trends where musicians sample existing music for inspiration to compose a new song. To a layman, it may seem like the score is copied. However, in the realm of music, it is completely legal to take inspiration from sampling if done ethically and the output is clearly different from the song it took inspiration from.[15]

Therefore, the defence of fair dealing as laid down by Justice T.V. Ramakrishanan in the landmark judgment of Civic Chandran v. Ammini Amma[16], can be used against musicians if they claim copyright over a certain melody created by A.I. having the same style. The case established that mere inspiration is not infringement if there is a visible difference in the output and the artistic work it took inspiration from. If an individual of ordinary prudence cannot differentiate the two, there stands no probability of infringement. This can be seen across multiple popular songs that were essentially sampled and no one knew for decades.[17]

This brings us to the second argument; the voice of the artist. As discussed, copyright does not exist in an idea but in its expression. An artist (in the present case a singer) has the right to manage their reputation as per Section 57 of the Copyright Act, 1957[18] that essentially aligns with the principles laid down in Article 6bis of the Berne Convention.[19]These provisions discuss moral rights. Moral rights grant creators of a work, the Right to Paternity and Integrity. Right to Paternity is essentially the right to claim authorship over one’s work. The Right to Integrity grants any author the right to object to any work that is prejudicial to their goodwill.  

In addition to moral rights, artists also have the economic right to receive compensation for any commercial gain by an unauthorised party. However, that exists in the works they have already created. When it comes to their voice (when not in expressed form), it is simply a part of their personality. While a singer can defend their personality rights specifically, they cannot claim copyright over their voice which is why there exists a legal lacuna in this field. This lacuna has essentially motivated multiple high-profile singers to sign petitions[20] and criticise A.I. as was previously mentioned.

Fundamental Challenges in Creating a Balance

The aforementioned discussions highlight the ambiguity in the existing law and the dire need for progressive reforms for the creative community concerning their rights in their work or personality. In a widely populated and developing country like India, technological advancement is an inevitable necessity. However, it is imperative that the promotion of growth does not render the creative depth and authenticity required for copyright of feeble importance. Consideration of artistic expression and the need for creativity in this very expression is necessary and should be diligently followed as has been the case since the onset of Intellectual Property Law and specifically in the domain of Copyright Law.

Copyright Law is built on the commercialisation of one’s creativity and labour in a tangible form.[21] It aims to ensure that every individual is duly compensated for work that utilised their labour and required creativity.[22] However, it is a difficult task to create a balanced and fair system of royalty/licensing rates in case artists sample their voices for machine learning or the training of A.I. source code, consensually. The creation of a fair system would require expert intervention and proper analysis which is time-consuming and can still be contested as subjective upon result.

Even if one ventures beyond the issue of copyright and towards an artist’s personality rights, it is the very personality, unique voice, and singing style that garners them reputation and goodwill. If these unique traits of the artist are machine-trained creating a multiplicity of works, their individuality as an artist and performer keeps on diminishing, consequently decreasing their face value in the industry.

Nonetheless, if a system or a technology is created that can make the transmission of these samples or the process of machine learning transparent and safe, there still exists a wide gap in the privacy of these samples. Companies would need to practice due diligence in storing, accessing and utilising this data. Lack of due diligence could lead to these samples being lost, hacked or taken by someone who is not authorised to access the same and the misuse of this data can cause grave loss of reputation, profits and unnecessary long litigations. There exists no law to safeguard this issue or any software that can guarantee to resolve this obstacle.

Possible Solutions and Way Forward

Rahman’s A.I. song brought a revolution in India in terms of music and its contemporary intersection with A.I. software.[23] He discussed how his team had taken due permission from the families before sampling the deceased artists’ voices as after their death, their personality rights were forwarded to their legal heirs and representatives.[24] It can be reasonably deduced that it is a utopian idea to believe that an imposition of a complete ban on the rising activity of A.I. across various creative fields is possible. It was simply a question of when it reached the artistic realms. Therefore, it is necessary to create a balance between the existing legal framework and the dynamic nature of the society.

To fill this bridge, it is essential to establish feasible royalty rates between artists and the A.I. companies sampling their voices to create new artistic work. Essentially, for the software to be proficient enough to create new renditions, its source code model has to be trained ethically, which would require optimal samples. This can be done in a legal-proof manner by formulating well-suited licensing agreements where musicians can license samples of their songs to train the A.I. models and in return receive due royalty for the samples distributed and commission in prospective profits from any song created using them. In cases, where the artists are deceased, the family and legal heirs should be compensated instead. The agreement would duly assess and state the rights and liabilities of both parties in a tangible format. Therefore in case of any dispute, the process of mediation or litigation is clear, efficient and completely ambiguity-free.

With technology rapidly developing, it is no shocker that creative realms are also welcoming machine learning. Areas which were considered completely creative are now seemingly competing against “machines” that were preconceived as technical and practical. Nevertheless, A.I., if welcomed foresightedly, understood carefully, and utilised ethically, can efficiently revolutionise the process of music creation and production in India. It can easily redefine the stringent boundaries of artistic expression as understood in India.

This highlights the importance of ensuring continuous discussion and discourse over the interplay of laws and society’s dynamic nature. Policy-makers and the Legislation have to stay on their toes and fill any lacuna that is created, especially, in a diverse, highly populated and culturally vibrant country like India. While the suggestions that have been provided by the author are not essentially tested, they are the only available mechanisms that can be put into effect till the Legislature comes up with an adequate solution to fill this ditch between safeguarding the rights of the creative community and not restricting the liberty of innovators to create.  

The evolving nature of human creativity in a technologically driven landscape can either be a boon or a bane. The decision, fortunately, rests in our hands.

[1] Rob Salkovitz, ‘Midjourney Founder David Holz on the Impact of AI on Art, Imagination and the Creative Economy’ Forbes (16 Sept 2022).

[2] U S Copyright Office, Zarya of the Dawn (Registration # VAu001480196, 2023).

[3] PTI, ‘A R Rahman on Using AI in Music: Not a Gimmick, has to Serve a Purpose’ The Hindu (3 Feb 2024); TOI, ‘AI in Music: A R Rahman's Latest Move to Recreate Voices of these Veteran Singers Creates Controversies’ The Times of India (5 Feb 2024); ET Online, ‘A R Rahman Recreates Voices of Two Late Music Icons for Rajinikanth's ‘Lal Salaam’ using AI, Fans Get Divided Over Ethical Implications’ The Economic Times (30 Jan 2024).

[4] Ibid.

[5] ET Online (n 3).

[6] David Friend, ‘Universal Music Group Pulls Down AI-Generated Song with Drake, The Weeknd’, (The Globe and Mail, 18 April 2023) <https://www.theglobeandmail.com/arts/music/article-universal-music-group-pulls-down-ai-generated-song-with-drake-the/> accessed 11 April 2024; Joe Coscarelli, ‘An AI Hit of Fake ‘Drake’ and ‘The Weeknd’ Rattles the Music World.’ The NY Times (24 Apr 2023).

[7] David Mouriquand, ‘Artists Sign Open Letter Warning against 'Predatory' Use of AI in Music’ Euronews, (3 Apr 2024); Tech Desk, ‘Katy Perry, Nicki Minaj Among 200 Artists Calling Out ‘Predatory’ AI In Music Industry’ The Indian Express (3 Apr 2024); Liv McMahon, ‘Billie Eilish and Nicki Minaj want stop to 'predatory' AI music’ BBC (2 Apr 2024).

[8] Hayleigh Bosher, ‘Artificial Intelligence is Proving to be a Legal Nightmare for the Music Industry’, (Scroll, 18 April 2023) <https://scroll.in/article/1047510/artificial-intelligence-is-proving-to-be-a-legal-nightmare-for-the-music-industry> accessed 11 April 2024.

[9] Bernard Marr, ‘Generative AI is Revolutionizing Music: the Vision for Democratizing Creation’ Forbes, (5 Oct 2023).

[10] Ibid.

[11] Arianna Johnson, ‘Spotify Removes ‘Tens of Thousands’ of AI-Generated Songs: Here’s why’ Forbes (9 Mar 2023); Ben Wodecki, ‘AI Covers Flood YouTube: Creativity or Controversy?’ (AI Business, 28 July 2023) <https://aibusiness.com/ml/ai-covers-flood-youtube-creativity-or-controversy-> accessed 12 April 2024.

[12] 2008 AIR SC 809.

[13] 1978 AIR SC 1613.

[14] Eastern Book Company (n 12).

[15] Ben Challis, ‘The Song Remains the Same: A Review of the Legalities of Music Sampling’ (WIPO Magazine, November 2009) <https://www.wipo.int/wipo_magazine/en/2009/06/article_0006.html> accessed 12 April 2024.

[16] 1996 SCC Online KER 63.

[17] Barnaby Lane, ‘15 Smash Hits You Might Not Have Known Sampled Other Songs’ (Business Insider, 12 April 2023) <https://www.businessinsider.com/hit-songs-that-use-samples-2023-4?IR=T> accessed 12 April 2024.

[18] Copyright Act s 57

[19] Berne Convention art 6bis.

[20] Mouriquand (n 7).

[21] Secretary to GOI, A Handbook of Copyright Law.

[22] Ibid.

[23] TOI (n 3).

[24] PTI (n 3).