Who Owns AI-Generated Music? I Asked People That Make It
Copyright laws need an update for our current day
If you’ve read about artificial intelligence and music in the popular press, you’ve likely heard of at least one company: Suno. A startup based in Massachusetts, Suno claims to be “building a future where anyone can make great music.” How? They allow you to enter a description into a text box — say, “afrocuban jazz song about cats” — and, in a matter of moments, they’ll return two full recordings that fit said description. Impressive? Yes. Somewhat frightening? Also, yes. Legal? That requires a longer discussion.
Who Owns AI-Generated Music? I Asked People That Make It
By Chris Dalla Riva
When Suno first launched to the public, there was speculation that the technology was built on thousands upon thousands of copyrighted songs. In August 2024, that speculation was no more. Suno admitted it when responding to a lawsuit filed by the world’s largest music labels:
It is no secret that the tens of millions of recordings that Suno’s model was trained on presumably included recordings whose rights are owned by the Plaintiffs in this case … Suno’s training data includes essentially all music files of reasonable quality that are accessible on the open Internet, abiding by paywalls, password protections, and the like, combined with similarly available text descriptions.
To be clear, the labels did not allege that Suno’s musical output is infringing on their copyrights. They alleged that you can’t use millions of copyrighted recordings to build a tool like this without permission from those who own the recordings. Suno, in response, claimed that this actually wasn’t an issue: “It is fair use under copyright law to make a copy of a protected work as part of a back-end technological process, invisible to the public, in the service of creating an ultimately non-infringing new product.”
Here’s a specific example of what they mean. When you prompt Suno to make a “50s rock n roll song about cars,” the labels can’t claim that the output infringes on their copyright because it sounds like a Chuck Berry song in their catalog. Styles of music are not copyrightable. James Brown, for example, cannot own funk music in its entirety despite how necessary he was in creating it. Similarly, Universal Music Group cannot own rock music despite their vast back-catalog of the genre’s foundational recordings.
Though Suno cites some interesting precedents where the intermediate use of copyrighted material was indeed deemed fair use, I’m not really convinced by their argument. There are no hard and fast rules about what constitutes fair use, but Section 107 of the Copyright Act lays out a framework that courts can work with.
Purpose and Character of Use: If a new work uses a copyrighted work in highly transformative way or is used for educational or non-profit purposes, then it is more likely to be deemed fair use.
Nature of the Copyrighted Work: If a new work uses a copyrighted work for commentary or criticism, then it is more likely to be deemed fair use.
Amount of Copyrighted Work Used in Relation to the Whole: If a new work only uses a small portion of a copyrighted work, then it is more likely to be deemed fair use.
Effect on Value of Original Copyrighted Work: If a new work is unlikely to compete with the copyrighted work it is using, then it is more likely to be deemed fair use.
Looking at these tenants, along with a few other considerations, it’s hard to imagine a judge looking favorably on Suno’s technology given their business model, the scope of copyrighted material that they’ve used, and the ability of their technology to devalue current music. While I could go on about this, let’s ask another group what they think, namely Suno’s users.
Talking to AI-Powered Artists
Most articles you read about Suno will feature arguments about copyright, conversations with established artists, or quotes from people who work at the company. Very few actually give any insight into the alleged millions of people using the platform. I decided to reach out to a few them.
ImOliver — that’s how he’s listed online — was the first person to have a song reach one million plays on Suno. Over email, he told me that he tried to learn an instrument in his teenage years but lacked the skills to bring his ideas to life. Suno changed that. “Suno,” he told me, “feels like an instrument that lets me focus on the creative side of writing while handling the technical work.” He went on, “It’s a shortcut to creating music that sounds good without years of training, which, honestly, at my age, I don’t have the time or the neuroplasticity for.”
When I asked about his creative process, he noted that while Suno helps with the arrangements, he considers the lyrics, emotion, and storytelling all his own. His goals, he relates to me, are lofty: “My goal is to become the first one-man company generating £1 billion using AI.”
MAdoK958, another Suno user that I spoke with, didn’t have plans quite as grand. He just wanted to make music that he liked. “[I]t’s a great bonus,” he wrote to me over Discord, “if others enjoy it too.” Nevertheless, MAdoK958, like ImOliver, never made music before Suno. He just fell in love with the technology after a friend showed it to him.
As you’d imagine, both ImOliver and MAdoK958 consider their usage of Suno a creative act. But do they feel they are the owners of the music that they prompt Suno to generate? Yes. In fact, MAdoK958 explicitly notes that each of the songs on his profile is owned by him (e.g., “Copyright © 2024 MAdoK958”). While he could understand why artists felt they should be compensated for their music being used to train Suno, he felt that Suno’s output was influenced by these inputs rather than copying them. “Should a rock artist pay their childhood favorite band just because their music was influenced by that band?” he asked me. “To me, that seems like a similar situation.”
ImOliver expressed similar sentiments on ownership but was even more skeptical of the notion that artists should be compensated for their songs being fed into the technology:
Artists shouldn’t be paid for AI-generated music unless the AI was explicitly trained to replicate their style or voice. Suno, like all creators, draws from collective influence to produce original compositions, much like how Oasis was inspired by The Beatles to create something new. Every musician has countless hidden influences, and Suno is similar—it learns, it doesn’t replicate.
The focus should be on transparency and ensuring AI isn’t used to exploit specific artists unfairly. If there is a payout, it’s unclear how much would directly benefit artists versus other stakeholders. It’s also worth noting that Universal Music Group (one of the plaintiffs) are building a music AI themselves with a company called Klay. There’s a danger that it's less about protecting artists and more about having a monopoly on AI-generated music.
For the record, I do agree with some of this. We do need to make sure that specific artists aren’t being exploited. We also should be wary of major labels just trying to sue everybody out of existence and then building the exact same technology. That said, I do think artists need to be compensated if their music is used as training data. Furthermore, they should be able to opt out of having their music used by these models.
No matter what side of this debate you fall on, I think we can agree that we need some sort of jurisprudential clarity. The world of AI-generated music is currently the wild west. We have artists and labels claiming that they own the output from these models because their music was used to build them. We have users on platforms claiming they own the output because they crafted a prompt or wrote some lyrics. We also have the platforms themselves claiming ownership. Here’s what Suno says about such in their terms of service:
By using the Service or otherwise transmitting Submissions to us, you grant to Suno … a worldwide, non-exclusive, fully paid-up, … royalty-free, perpetual, irrevocable right and license to use, reproduce, store, modify, distribute, create derivative works based on, perform, display, communicate, transmit and otherwise make available any and all Content (in whole or in part) in any media now known or hereafter developed, in connection with the provision, use, monetization, promotion, marketing, and improvement of our products and services, including the Service and the artificial intelligence and machine learning models related to the Service.
You read that right. Suno is claiming that they can use anything you generate on their platform now and forever after in any medium, even if that medium does not currently exist. That is, in a word, crazy. And it suggests that there needs to be an in depth clarification of how musical copyright works in this space.
To be clear, musical copyright is not some abstract idea handed down from the heavens. It has evolved over centuries. In the United States, musical compositions were first granted intellectual property right protections in 1831. In 1897, protections were extended to performances of songs. In 1909, they were extended to mechanical reproductions of songs. In 1972, they were extended to recordings. It’s time for us to explicitly define these rights in the world of AI. Whether you like it or not, this technology is not going away.
A New One
"One Take" by Smileyy
2024 - Melodic Rap
“One Take” is the name of Smileyy’s latest single, but it’s likely a good descriptor too. Not even two minutes long, Smileyy’s laid back rhymes are some of the most nonchalant I’ve heard in recent memory. That doesn’t mean they aren’t dexterous, though. It just sounds like that dexterity comes easy to the Georgia rapper. In fact, if “One Take” were recorded in anything more than one take, I would be shocked.
An Old One
"Be Young, Be Foolish, Be Happy" by The Tams
1968 - Beach Music
If you are ever trying to find forgotten musical gems, just turn to any book on the history of dance music. You’ll likely read about some gorgeous song that was played on a dance floor incessantly in the 1960s and 1970s and was then lost to time. The Tams’ soulful cut “Be Young, Be Foolish, Be Happy” is a perfect example of what I’m describing.
A minor hit when it was first released in 1968, “Be Young, Be Foolish, Be Happy” caught a second wind when British DJs began spinning it in the northern soul scene of the early 1970s. The song would have never made it to my ears if I didn’t read about it in Last Night a DJ Saved My Life, an exhaustive history of the disc jockey by Bill Brewster and Frank Broughton. This song alone was worth the book’s 500-page price of admission.
Shout out to the paid subscribers who allow this newsletter to exist. Along with getting access to our entire archive, subscribers unlock biweekly interviews with people driving the music industry, monthly round-ups of the most important stories in music, and priority when submitting questions for our mailbag. Consider becoming a paid subscriber today!
Recent Paid Subscriber Interviews: Pitchfork’s Editor-in-Chief • Classical Pianist • Spotify’s Former Data Guru • Music Supervisor • John Legend Collaborator • Wedding DJ • What It’s Like to Go Viral • Adele Collaborator
Recent Newsletters: Personalized Lies • The Lostwave Story • Blockbuster Nostalgia • Weird Band Names • Recorded Music is a Hoax • A Frank Sinatra Mystery • The 40-Year Test
Want to hear the music that I make? Check out my latest single “Overloving” wherever you stream music.
There is only one jazz song about cats that anyone ever needs and no AI could ever match it: https://youtu.be/RZq3_mASEbo?si=yG0ZSMLs_m-LZgDM
This ImOliver chap is not interested in music, he is interested in money. There's a difference.
Nice new legal strategem but I'm suing the equity company instead for this mess.