[trIPping] “Artist X type of song”: should the (unauthorized) use of voices be allowed in AI-produced demos?
[this is trIPing, where we discuss the craziest ideas and possibilities in Intellectual Property. No final thoughts, no legal advice, just tripping]
Since there’s no escaping the spread of AI in the music industry, which invariably includes the use of multiple IPs and rights, we might as well define when that’s acceptable or not.
The laws for using copyrighted material and elements protected by other rights are all set already. Until now, there are no limitations or exceptions made for AI. But the question: should there be?
I’ve been thinking about what this means for music production and songwriting.
For a songwriter or producer, it sure helps when a demo gets a close feel of what the song will sound like if recorded by the artist you’re pitching it to.
But unless you’re like Russ, producing such a high-quality demo demands being a part of a wildly talented team that is willing to do it all for free, or an extraordinary amount of money to afford the costs of production, recording, and editing vocals, mixing, and mastering — all that just for a demo.
In the K-pop industry, for example, the closer a demo sounds to an actual track released by a K-pop group, the bigger the chances that it makes the final cut for a K-pop album. That’s understandable but that’s also why the K-pop songwriter dream is such an expensive one, since the bar for demos is so high.
Could AI help reduce these costs? And what if AI could help build the bridge between what the songwriter envisions and what the A&R needs?
Putting aside discussions on whether AI-composed music sucks or not, AI can do a lot to help people who actually write good songs. Some of these people just need a little push to help everyone see the potential their songs have.
And because sometimes people will not be convinced until they actually hear it, AI can be a shortcut to create the “realest” demos. But is that legally allowed? And if it’s not, what if it was?
Let’s explore this possibility, but first, let’s understand what it would mean through the lens of a recent controversy involving one of the world’s biggest artists.
The Bad Bunny x FlowGPT case
On October 2023, when Bad Bunny released Nadie Sabe lo que va Pasar Mañana, fans were surprised to hear the album had little reggaeton (the genre that made his previous album Un Verano sin Ti break several records).
Days later, after the announcement of his next world tour, the Puerto Rican star let fans know through his official WhatsApp channel that this tour was not made for the fans of his reggaeton songs. “Only trap! In this tour you’re not going to hear ‘Ojitos Lindos’ nor ‘Moscow Mule’… You’re going to hear ‘Mr. October’, ‘Telefono Nuevo’ and all that”.
Except that Bad Bunny fans’ thirst for new reggaeton songs was satiated in another way.
Just a few weeks after Nadie Sabe, AI-producer FlowGPT released “Demo #5: Nostalgia”, an original composition featuring “fake” vocals of Bad Bunny as well as Justin Bieber and Daddy Yankee.
Many Bad Bunny fans enjoyed “Demo #5: Nostalgia” and the track was blowing up on TikTok and doing great numbers on Spotify before it was removed.
But Bad Bunny was not pleased.
“If you like this shit song that’s going viral on TikTok then get out of this group chat right now”, he said on WhatsApp. “You don’t deserve to be my friends and that’s why I made this new album, to get rid of people like you. And I don’t want you on the tour either.”
There are several legal layers in this case.
First, of course, the song is an unauthorized use of Bad Bunny’s voice (through manipulation of actual recordings of his voice) and an infringement of his publicity rights.
But also, the track and its dissemination goes against Bad Bunny’s current artistic project. That can be a violation of his moral rights since the artist does not want to be associated with what’s being done with his name, image, or voice. That’s easier to understand in countries that acknowledge moral rights, such as Brazil.
It’s an awkward situation. But let’s say “Demo #5: Nostalgia” was really just a demo that FlowGPT pitched to Bad Bunny’s team and either was accepted and recorded by Bad Bunny or just never saw the light of the sun. Would that’ve been less of a problem?
Can authorization be excused for experimentation, such as in music demos?
The case for copyrights
When it comes to copyrighted material, copyright laws already feature limitations on the rule of asking for the copyright holder’s permission to use their work.
In the U.S., for example, fair use may be enough to handle disputes regarding the unauthorized use of melodies, lyrics, or recorded tracks, whether it’s done through AI or not.
There are 4 factors to go through when examining fair use: (1) the purpose and character of the use; (2) the nature of the copyrighted work; (3) the amount or substantiality of the portion used; (4) the effect of the use on the potential market for or value of the work.
If a judge understands your work meets all the requirements to qualify for fair use, then you’re cool.
When it comes to demos or overall tracks produced just for the sake of experimentation, there have been decisions on using uncleared samples qualifying as fair use. That was the case of Nicki Minaj using an unauthorized Tracy Chapman sample just for the means of seeing how it’d sound to then try clearing the sample (even though later they ended up settling).
The problem was that Minaj ended up playing the demo on the radio. Should it stayed just as a demo that was never played outside the studio, Nicki would’ve been fine.
As for now, there’s nothing stating that fair use can’t apply to AI-music. So at least in theory, it’s safe to say the fair use stakes are the same in music written or produced through AI and in music 100% made by humans.
But things can get a little more complicated when it comes to AI that manipulates singers’ voices.
The case for personality rights
Voice is not protected by copyrights. The unauthorized use of a voice may result in several legal outputs, such as unfair competition, and violation of personality rights/rights of publicity.
Some authors defend the possibility of “applying a modified fair use doctrine to certain right of publicity cases” (Randall T.E. Coyne, Toward a Modified Fair Use Defense in Right of Publicity Cases, 29 Wm. & Mary L. Rev. 781 (1988), https://scholarship.law.wm.edu/wmlr/vol29/iss4/4 — disclaimer: Coyne’s article does not discuss this theory within the realm of AI).
If fair use applied to the use of images, voices, and other moral attributes, then “experimenting” with these assets would not qualify as infringement — as long as it remains an experiment and does not bring actual damage to the rights holder, of course.
“When the alleged infringement involves the appropriation of plaintiff’s right of performance, the use is presumptively unfair. Because such a use directly reduces the market for plaintiff’s work, it places a strong economic disincentive in the path of plaintiff’s creative endeavor”
(Randall T.E. Coyne, Toward a Modified Fair Use Defense in Right of Publicity Cases)
Think of the beatmakers who “appropriate” famous artists’ “styles”, and use their names to attract creatives who are interested in following such artists’ steps, putting out “Artist X type of beats” tracks and videos out there.
IP-wise, stealing a style or genre is hardly possible. And sometimes, not even the artists are offended — in some cases, they even like it. J. Cole liked one “J. Cole type of beat” so much that he even used it.
In a similar way, would it qualify as fair use if an artist’s voice is manipulated for a demo, just for the sake of showing if a track will fit said artist’s style?
One may argue that this would still be unfair competition since it misappropriates a series of creative elements that are the outcome of continuous efforts and investments. But it is also true that to enable cultural movements, stating more permissive bars for Intellectual Property laws is just as important as paying what’s due to the rights holders.
Being a trendsetter and a trailblazing artist has its bonus and its onus. Ultimately, finding a harmonious equilibrium between AI-driven innovation and artistic integrity will be crucial for the continued evolution of the music industry.
But since we can hardly prohibit the use of AI per se, maybe it’s time we define what battles are worth fighting for.
trIPing aims to be a fun and relaxed space to discuss IP theories and possibilities. Our thoughts are based on the laws in force at the time it was written, but please do not use this post as legal advice. Seek a lawyer.
This blog post does not create an attorney-client relationship, nor replaces professional legal services or business consultancy services.
3Três Consultoria e Criação (Consulting & Creative) specializes in the intersection of creative, Intellectual Property Law, and communication services.
Contact us at: contato@3trescc.com
