United States: Embracing AI's Disruption Of The Music Industry: Advantages, Challenges And The Future

If alarm bells were already ringing in the music business, they grew deafening when the AI-generated track “Heart on My Sleeve” made headlines in April of 2023. The now-infamous song harnessed AI voice-cloning technology to create a startlingly lifelike — but ultimately “deepfake” — collaborative performance between artists Drake and The Weeknd.

When the song hit streaming services, it garnered more than half a million streams before Universal Music Group (”UMG”) raced to have it taken down. In a somewhat dramatic statement, UMG asked the stark question: “which side of history [do] stakeholders in the music ecosystem want to be on: the side of artists, fans, and human creative expression, or on the side of deep fakes, fraud, and denying artists their due compensation?”

This question — admittedly uttered at a time when labels and publishers had been sent into a sudden tailspin — establishes what some might call a false dichotomy. The idea that the introduction of AI technology to the music industry must create an absolute division between creative expression on one side, and fraud on the other, would seem to suppress, rather than promote, the harnessing of available tools to enhance artistry and innovation. 

After all, new technologies have constantly rubbed up against the music industry — from the introduction of vinyl records to the advent of digital streaming services — and, ultimately, have been assimilated and folded into industry norms, rather than rejected. The same goes for the inclusion of musical tools and technologies such as sampling and synthesizers. Why should AI create such fear and apprehension in the industry when it poses a similar potential to enhance creative expression, rather than undermine?

While “Heart on My Sleeve” incorporated unabashed “deepfakes,” consider the alternative use of AI to promote pure “human creative expression” — the alternative side of UMG's dichotomy. Only two months after the Drake shake-up, the Beatles announced that they had used AI to “extricate” John Lennon's voice from an old demo to complete a final song. When it came to making what McCartney called “the last Beatles record,” he said: “we were able to take John's voice and get it pure through this AI. So then, we could mix the record as you would normally do.… So, there's a good side to it and a scary side — and we'll just have to see where that leads.” 

This time, the record was met with critical acclaim; an instance showcasing the use of AI technologies as a tool that could create a unique piece of art harnessing the voice of an original band member, rather than, as UMG fears, an unauthorized means to clone artists and deny them royalties they were owed. Turning away from the use of AI for cloning and copying (the “deepfake” pathway), the Beatles example opens the door to a second pathway — the use of AI for creating truly original music. AI, it seems, can be part of both sides of the coin.

The “deepfake” pathway, of course, assuming that the uses of talent and celebrity voices and other attributes are unlicensed, has raised controversy in the industry from both a royalties and IP infringement perspective. That said, brands and consumers alike are jumping on the deepfake bandwagon to create viral content, and it is quickly evolving into a widespread practice.

The “original music” pathway, on the other hand, is subject to much more creative control by the artist — but still implicates the inherent risks of using AI technology and outputs, based on an unknowable universe of training data, and the PR risks of public scrutiny as to the degree of creativity involved.

That said, both pathways potentially involve some degree of inherent creativity — from human judgment in writing, arranging and recording underlying musical elements to a track and then overlaying and adjusting a synthesized artist's voice to mesh with the overall production — to true composition and incorporation of isolated AI elements into entirely new content. Neither should be dismissed as a vehicle for accelerating and accentuating human creativity in expanding the musical universe.

Skill v. pushing a button: ‘Democratizing music' or undermining artists?

All this is not to say that AI will magically even the playing field with the click of a button — putting inexperienced creators online on the same level as musically inclined, sophisticated artists creating new compositions. Like the use of any technology or program, content creators will need to learn how to use AI skillfully, tweaking prompts, inputs, and the use of underlying training data, deep learning and reinforcement to generate better and higher quality content.

Rather than truly “democratizing” the music industry by creating a lower barrier to achieve the same results, AI could also be viewed as simply opening the door to a greater universe of musical experts as well as non-musical newcomers to generate exciting results (or, as time will tell, perhaps not).

Given the vast spectrum of results that could be achieved depending on the vision and skill of the user, AI could very well become the great musical “equalizer” — allowing participation by all without undermining the talent of existing artists. Only industry shifts will determine the trajectory and impact of AI on musical composition, but one thing is for sure — it isn't going anywhere soon.

Given the potential for skillful — and creative — use of AI, it is somewhat curious that the industry's knee-jerk reaction has initially been to cut off the inherent merits of non-human creations at the pass. The Grammys was one of the first industry groups to state that AI-generated works would not be eligible for awards on their own merit — an approach that mirrored the Copyright Office's guidance that it would not grant copyrights to “works that lack human authorship” and that “the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author.”

Playing out this position, the Copyright Office declined to recognize copyright protection in Stephen Thaler's AI-generated image, “A Recent Entrance to Paradise,” on the grounds that the image “lacked the required human authorship necessary to sustain a claim in copyright,” because Thaler had “provided no evidence on sufficient creative input or intervention by a human author in the Work” — even though Thaler used his own AI algorithm to generate the output.

The Copyright Office also stated that it would not “abandon its longstanding interpretation of the Copyright Act, Supreme Court, and lower court judicial precedent that a work meets the legal and formal requirements of copyright protection only if it is created by a human author.”

The Copyright Office's guidance and follow-up circulars do suggest that going forward, it will evaluate the degree of meaningful creative input from a human author when determining whether a given work (or a portion thereof) is eligible for copyright protection.

For example, the use of AI to sharpen a human-created image (e.g., by improving definition or color) may not undermine the protectability of the underlying image, but the use of AI technologies to generate an image based solely on training data, notwithstanding the quality of the prompts, may.

The Grammys seems to be taking a hint from this follow-up guidance, implementing a change to the ceremony's guidelines in June. The president stated that “AI, or music that contains AI-created elements, is absolutely eligible for entry and for consideration for Grammy nomination… as long as the human is contributing in a more than de minimis amount, which to us means a meaningful way, they are and will always be considered for a nomination or a win. What's not going to happen is we are not going to give a Grammy or Grammy nomination to the AI portion.”

We are at a pivotal moment to see how these positions will play out in practice — especially since the use of AI to enhance, improve or sharpen human-created works is likely to be a non-severable feature core to the work. Take the Beatles example: does the use of the John Lennon track undermine the inherent creativity involved in the song composition? If the track were nominated for a Grammy, how would it play out in practice if an award could not be awarded to the “AI portion” — i.e., John Lennon's voice? After all, is this not simply the modern-day equivalent of the virtually edited “Unforgettable” duet between Natalie Cole and her deceased father, Nat King Cole — which indeed won several Grammys in 1992?

Ultimately, it seems likely that AI will be incorporated into music akin to the way special effects are incorporated in cinema — from enhancing images to de-aging, special effects and other techniques requiring human judgment to deploy — and ultimately will be a part of the holistic creative endeavor. As such, the Grammys position seems somewhat short-sighted.

It is likely that as AI technologies become industry practice in the same manner as special effects and sound enhancers, we will see a greater push towards treating AI as an integral, rather than severable, feature of a given composition — if not with respect to the entire composition itself.

Even if worthy of merit — what about IP rights?

All this being said, we cannot overlook the intellectual property and rights issues implicated by the use of AI technologies — especially when using copyrighted works as part of training data (and as a result, potentially being incorporated or replicated in the output) or harnessing or cloning third party talent without the appropriate licenses.

By way of analogy, several novelists and comedians — most recently Sarah Silverman — have brought copyright actions against AI platforms that have “scraped” their copyrighted works to assimilate into their training data, while labels and artists alike have emphasized the need for artists to receive a royalty or revenue share for the use of their talent in AI-generated works. Recall the early days of music “sampling” — initially viewed by the industry as simply artist interpolation and “fair use,” but ultimately amounting to unauthorized content that needed to be licensed.

The ostensible position of many AI platforms is that the “scraping” itself should be considered an operational fair use — the digital equivalent of a human being watching thousands of YouTube clips of musicians and transcriptions to “train” themselves, or to serve as “inspiration,” such that the only issue is related to whether the AI output itself is so substantially similar as to infringe on the underlying work.

However, the artist industry is pushing back — especially when AI platforms use platforms such as torrents to scrape copyrighted works such as novels that are otherwise not publicly available on the internet.

The U.S. Supreme Court recently further analyzed the fair use argument in the context of the “transformative” use of existing material for a new artistic piece, addressing whether Warhol's “Orange Prince” silkscreen, licensed to Conde Nast, was sufficiently transformative of the celebrity photographer Lynn Goldsmith's original portrait of Prince to qualify as fair use. The Court ruled in a 7-2 decision that the Warhol Foundation's use does not weigh in favor of the Foundation under the first factor of the fair use defense to copyright infringement (Andy Warhol Foundation for the Visual Arts Inc. v. Goldsmith, 143 S. Ct. 1258 (2023)).

In particular, the Court rejected the Foundation's argument that the use was “transformative,” since in the Court's view the “purpose and character” of the use was for substantially the same commercial purpose as the underlying work; according to the Court, “Goldsmith's photograph and AWF's 2016 licensing of Orange Prince share substantially the same purpose” because “AWF's use of Goldsmith's photo was of a commercial nature — [which] counsel[s] against fair use, absent some other justification for copying.” 

The case, one of the first Supreme Court decisions to consider the fair use argument in several years, is likely to play out in the context of the proliferation of AI-generated content — perhaps to the benefit of visual artists whose artworks have been used in the training of AI systems, if the resultant AI output is used for a substantially similar purpose to the purpose of the original artwork.

The key moving forward will be to look closely at how the AI output utilizing an underlying work is being used — in addition to the analysis of whether the output is substantially similar to it or otherwise uses unauthorized elements. Perhaps this is why platforms such as OpenAI are moving towards actually partnering with content platforms to license their material for use as training data — with an associated monetary benefit or royalty share if incorporated into the output.

How could AI positively impact the music industry?

We are at an inflection point with respect to the harnessing of AI in the music industry — a time when artists and creators face a choice to fear the use of AI, to use it in a way that risks claims by third parties, or to embrace it head-on.

The latter is not a new strategy — as early on as 2019, musical professionals such as Holly Herndon have encouraged users' use of open AI tools to collaborate directly with artists. For example, Holly's AI deepfake “twin,” Holly+, is a website where users can upload any polyphonic audio and have it transformed into a download of music sung in Herndon's voice.

The year 2023 has seen artists such as Grimes hopping on the bandwagon to welcome AI into their musical approach, in Grimes' case, encouraging fans to make music using her AI-generated voice through her website Elf.Tech.

In a bid to normalize revenue-sharing for such collaborative projects, Grimes went a step further and promised that any fans making AI music using her voice-clone would be able to share in the profits from any project they create, tweeting: “I'll split 50% royalties on any successful AI generated song that uses my voice. Same deal as I would with any artist I collab with. Feel free to use my voice without penalty. I have no label and no legal bindings. I think it's cool to be fused with a machine and I like the idea of open sourcing all art.”

In the wake of this movement, and ostensibly in an attempt to get ahead of unlicensed uses, UMG has started actively encouraging its artists to adopt voice-cloning technology, signing a deal with AI startup Endel, a “first of its kind strategic relationship” that enables artists to create “science-backed” soundscapes designed to “enhance listeners' wellness.”

On the flip side, other artists are taking a riskier approach, emphasizing the creative (but not necessarily commercial) benefits of using AI deepfake technologies. For example, David Guetta recently released a video during one of his live sets that used AI technology to add a cloned voice of Eminem to one of his songs.

In the caption, he wrote: “Eminem bro, there's something that I made as a joke and it works so good — I could not believe it! … Basically you can write lyrics in the style of any artist you like, so I typed: ‘write a verse in the style of Eminem about future rave,' and I went to another AI website that can recreate the voice. I put the text in that and I played the record and people went nuts… obviously, I won't be releasing this commercially.”

While the video went viral and Eminem has yet to complain publicly, the release sparked a virulent debate online regarding Guetta's legal and ethical obligations to Eminem — as well as to the public — when harnessing such a “deepfake.”

Power, responsibility, and embracing disruption

Underpinning the legal framework for the use of AI in music is a broader consideration — that of the responsible and transparent use of AI technologies. For example, what if David Guetta had not disclosed to the public that the use of Eminem's voice was a deepfake? This raises a separate issue — that of authenticity, ethical responsibility to the target demographic of consumers, and transparency of creation. 

Ultimately, this transparency is key; even federal regulators at the FTC have suggested that the failure to disclose the incorporation of AI technologies into a creative work (or even the use of AI to generate entirely new work) could be considered a deceptive “dark pattern” that ultimately misleads consumers.

While their viewpoint may continue to evolve as the technology does (since ostensibly, the use of AI to generate entirely new material may not necessarily be deceptive in the absence of “deepfakes” or other violations of third-party rights), artists and creators would do well to keep this in mind; noting that at the end of the day, AI is not a replacement for human creativity but rather a tool to enhance it, and the disclosure of its use — taking into account the legal guardrails — is not something to be feared but rather embraced. 

It is clear that AI's integration into the music industry holds immense promise, but it also requires careful consideration and responsible, ethical and, when appropriate, transparent use. As we navigate the potential challenges and opportunities presented by AI-generated music, a balance must be struck between technological advancement and preserving the authenticity and creativity of human artists.

By embracing AI as a collaborative tool within the appropriate framework and parameters, musicians have a unique opportunity to shape the future of music and provide novel experiences to audiences while maintaining ethical practices and respecting intellectual property rights.

Originally Published by Westlaw Today, 10 August 2023

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.