It’s Lucy and the football time with the music industry. Having led artists into a complete disaster with streaming, they seem poised to do it again, on a grander scale, with generative AI. Warner has a deal with Suno today, following Udio earlier this week. Peace for our time.
If only there were some kind of seconds of footage that conveyed the omnious feeling I have about this. Oh, yeah…
Warner Music Group’s CEO Robert Kyncl calls his Suno-WMG pact a “victory for the creative community that benefits everyone.” (Wait, everyone? Great, I’ll watch my bank balance.)
Under the agreement:
- Warner drops their Suno lawsuit
- Suno says they’ll now be able to “build a new generation of Suno models using high-quality licensed music.”
- Paid Suno accounts will be required for downloads (presumably to pay WMG statutory licensing fees)
- WMG artists will, on an opt-in basis, be able to provide names, images, likenesses, voices, and compositions for songs
Note that it does not appear that use in training data is opt-in; since Warner dropped their lawsuit, Suno is free to continue pillaging WMG-signed artists’ music to train its generative models.
The discussion of the technology asides, it’s Suno’s offering — focused on getting people to generate as much “content” as possible, drawing as heavily on the data set as it can — that’s a worry. It’s a normalized generative model built on artist work that goes uncredited, almost purpose built to produce the kind of slop in music that we’ve gotten in imagery. And it comes from a company that’s on the record with a consistently anti-artist message. Warner was one of the players that was big enough to go toe to toe, and they’ve capitulated.
And you can see why, like streaming, this is anything but “democratizing” as a force. Big players, bigger deals, more big platforms forcing the larger industry down everyone’s throats. I mean, if anything, artists might be better advised if they did want to do this to build their own AI models.
And this all sounds eerily familiar.
Good grief
See, the industry is probably thinking about Napster, which peaked early in 2001 but ultimately fell victim to lawsuits. Doing a deal with Suno does suggest the opportunity to find shared revenue while Suno is hot, rather than taking it down in lawsuits.
With Spotify, the relationship was again troubled — but only at first. WMG did its deal with Spotify in 2017, which cleared the way for Spotify’s public listing. And WMG was the last big holdout at the time, too, following Universal, Merlin, and Sony. Earlier in 2025, even, they doubled down, inking a new multi-year deal that they promise will yield greater revenues.
But in terms of the greater value of music, we have reason to be infuriated. Sorry, I mean “gently skeptical.” Let’s compare statements. Here’s Warner’s CDO in 2017 (then, Ole Obermann) talking about Spotify:
“We’ve arrived at a balanced set of future-focused deal terms. Together with Spotify, we’ve found inventive ways to reinforce the value of music, create additional benefits for artists, and excite their fans all over the world. Even with the current pace of growth, there’s still so much potential for music subscription to reach new audiences and territories.”
And over eight years later, this week’s statements by Kyncl:
With Suno rapidly scaling, both in users and monetization, we’ve seized this opportunity to shape models that expand revenue and deliver new fan experiences. AI becomes pro-artist when it adheres to our principles.
And the Udio statement:
We’re unwaveringly committed to the protection of the rights of our artists and songwriters, and Udio has taken meaningful steps to ensure that the music on its service will be authorized and licensed. This collaboration aligns with our broader efforts to responsibly unlock AI’s potential – fueling new creative and commercial possibilities while continuing to deliver innovative experiences for fans.
(There was also a deal with KLAY around licensing.)
Of course, if you believe the music industry is out there to protect artists’ IP rights and maximize the flow of revenue to you, you haven’t read up on history (or, possibly, your contract). But that’s another story.
Anyway, on the Udio side, we even have a little more of a vision of what this all might look like. Just as streaming hoped to get ahead of pirated downloads, now the pitch from licensed AI is to make sure that the industry can control and profit from you cloning your favorite artists. The bad news is that for anyone who, you know, values human beings as individual flesh-and-blood entities rather than data stores for generative algorithms to mine, your label is going to be encouraging fans to clone you:
Udio’s reimagined subscription service will introduce a suite of creative experiences that enable users to make remixes, covers, and new songs using the voices of artists and compositions of songwriters who choose to participate, while ensuring artists and songwriters are credited and paid. Ahead of the launch, Udio will also be rolling out expanded protections and other measures designed to safeguard the rights of artists and songwriters.
Uh, great. Wait, I mean, couldn’t you already make remixes and, like, sing songs to make covers? Are we really just giving up on production and voices?
I don’t want to be totally reductive here; I think there’s great creative potential in voice models for vocal synths — basically, the ability to produce your own Hatsune Miku. But even that example shows a contrast: it was creating a single character, that character was fictional, and a lot of human-controlled synthesis lies behind it. We know that the idea here is to eliminate all the actual human activity from the creative process — no more fans singing songs themselves on TikTok, and not a tailored instrument based around a particular vocal profile (as a sample library or standalone AI-based voice might do). Just more megacorporations doing deals with each other.
And if you thought Warner was going to fight for your music to not get gobbled up into a training set, you thought wrong.
Let’s just insult musicians
It’s not enough for Suno to ingest massive amounts of recorded music. They also apparently need to undermine the value of what music-making is. How else do you explain the tragedy of their own “Head of Creators” saying this?
These things don’t even match up. The entire world of music recording is built on demos, like, on tapes. (Isn’t the device she’s singing into a microphone?) What is she even talking about? Just before we all get sucked into the gaslighting here, she’s describing a proprietary text-based generative system that spews out generic music beds and requires a paid subscription of up to about $24 per month. And like all such subscriptions, those costs are expected to go up. Contrast the many free options available in both 2006 and 2025.
How do you square this lofty argument about unlocking access with a startup that has accumulated $2.45 billion in market value and hundreds of millions in funding? United Musicians & Allied Workers (UMAW) has exactly the right response to this.
It’s an insult. It’s an insult to kids, it’s an insult to musicians, and it’s an insult to the many people who have worked to provide actual access to music education, skills, and technology, often with volunteer labor.
And as I said, we’ve heard this before. Streaming was supposed to have the same magical impact. Yet in both cases, the people talking about democracy and access are inking multibillion-dollar deals for themselves.
The problem as always isn’t the tech. It’s big, unregulated capital.
Wasn’t the world on fire or something?
Then there’s the environmental impact. Turning this into a service, as opposed to just investing to train an individual vocal synth, for instance, sounds catastrophic. Listen, not all use of machine learning is comparable, and not all of it is particularly inefficient or environmentally unfriendly. But the kind of generative AI tech used by Suno has the potential to be fairly significant. (See mureka.ai’s study of music, specifically.) Various combined generative use cases are adding up, adding pressure on energy grids and water consumption, right when we’re supposed to be making massive cuts.
MIT Technology Report did a decent write-up and video if you’re pressed for time. (One apology, as a US citizen: the typical USA approach here is to scale up from the individual to a universe that ends with the continental 48 United States. But the points remain. They’re worse, in fact, looking at global impact and the ways things like water consumption amplify social injustices.)
The microwave is also a slightly US-centric image, but the metaphor holds: imagine leaving a major appliance on for an hour or hours.
So, remember this? (Just while I’m shaming WMG press releases, why stop, really?)
WARNER MUSIC GROUP RELEASES 2023 ANNUAL ESG IMPACT REPORT
There should have already been a red flag in the phrase “reducing WMG’s negative impact on the environment,” as that report also mentions their supposedly progressive stance on AI, while failing to mention AI’s outsized environmental impact.
And all of that aside, this is the same Suno whose CEO said this — let me just quote the amazing Joshua Eustis:
I think the majority of us don’t enjoy the entirety of the time we spend listening to Mikey Shulman saying words. As I’ve noted before, it does sound good if Anakin says it. (Sorry, AI.)
But I’m completely optimistic
About this deal? Ha, no, completely not. (Suno also got Songkick for some reason as part of the new marriage.)
No, I’m optimistic because I think the products in this case are basically crap (longer story, but because they limit you to producing what amounts to sound-alike stock clips), the opportunistic tie-ins for everyone’s identity are likely to be unappealing to fans and artists alike, and in the end, this will all be history. Music making has survived for tens of thousands of years — possibly even outside our immediate species — and will outlive this hideous bubble and the gaslighting from the people trying to prop it up. There are interesting applications for machine learning tools, just as there have been generative and algorithmic concepts for generations and neural net applications for well over a half century. They don’t all require massing hordes of data sets to commodify music just so a handful of people can get wealthy at everyone else’s expense.
If only there were something stupidly obvious from Warner’s past that fit the situation.