AI Music’s Copyright War Just Got Bigger: Labels Target YouTube “Stream-Ripping,” Suno and Udio

AI Music’s Copyright War Just Got Bigger: Labels Target YouTube “Stream-Ripping,” Suno and Udio

The battle over AI-generated music is no longer just about whether copyrighted songs can be used to train an artificial-intelligence model. Now the fight is moving deeper into how those songs were allegedly obtained in the first place.

Universal Music Group and Sony Music Entertainment have expanded their case against Suno with allegations involving audio obtained from YouTube. A federal judge allowed the labels to add a claim under the Digital Millennium Copyright Act accusing Suno of circumventing YouTube's anti-downloading protections. Suno later acknowledged in a court filing that it obtained YouTube audio for training using YT-DLP, while disputing the labels' legal theory and maintaining defenses including fair use.

FROM AI TRAINING TO “STREAM-RIPPING”

This distinction could become extremely important. For years, one of the biggest unanswered questions surrounding generative AI has been whether training an AI system on copyrighted material without permission can qualify as fair use. But the labels are now pushing another question: how did the AI company get the music?

The labels allege Suno circumvented technological measures used by YouTube to control access to downloadable copies. The court allowed Universal and Sony to pursue that DMCA claim, although allowing the claim to proceed is not a ruling that Suno actually violated the law.

Udio has faced a similar accusation. In its litigation with Sony, Udio acknowledged obtaining YouTube audio for training and using YT-DLP for at least some of that material. Udio continues to argue that training its AI is fair use.

That creates two separate questions that could shape the future of AI music: was using copyrighted music to train the model lawful? And even if training could qualify as fair use, was the method used to acquire that music lawful? Those aren't necessarily the same question.

THE MAJOR LABELS AREN'T FOLLOWING THE SAME STRATEGY ANYMORE

What started as a relatively united industry attack on AI music has evolved into a combination of lawsuits, settlements and licensing agreements.

Warner Music Group settled its litigation with Suno and entered a licensing relationship with the company. Universal, meanwhile, continues its litigation against Suno alongside Sony. Udio's situation is different again. Universal and Warner settled with Udio and moved toward licensed AI-music arrangements. Sony did not and remains in litigation with Udio.

So the industry's response to AI is becoming much more complicated than simply record labels versus artificial intelligence. Some companies are fighting AI platforms in court while simultaneously building licensing systems with other AI companies.

SONY IS STILL FIGHTING UDIO

Sony dramatically escalated its Udio battle by filing another lawsuit involving 30,117 sound recordings. Sony says those recordings were copied without authorization for AI training.

The lawsuit followed a procedural decision preventing Sony from adding more than 30,000 recordings to its original case. Rather than abandoning those claims, Sony filed another action.

The numbers illustrate how quickly these cases can become financially massive. U.S. copyright law can allow statutory damages of as much as $150,000 per work for willful infringement, although that represents a legal maximum—not a prediction of what Sony would ultimately receive.

THEN CAME SUNO V6

Universal and Sony filed another lawsuit against Suno involving 60,202 sound recordings. The labels' argument goes beyond simply claiming that an older AI system was trained improperly. They're attacking the lineage of Suno's newer technology.

According to the complaint, the labels contend that a newer model cannot necessarily escape problems associated with an earlier model simply by being rebuilt or distilled through another generation of technology. They characterize Suno's newer v6 technology as stemming from the same allegedly infringing foundation.

At the maximum statutory rate for willful infringement, 60,202 works could theoretically represent more than $9 billion in exposure. Again, that's a mathematical ceiling based on the statute—not an assessment of what a court would actually award.

This could become one of the most consequential issues in generative AI: can a company create a legally “clean” AI model using knowledge derived from an earlier model whose training is being challenged? That question reaches far beyond music.

ARTISTS ARE FIGHTING THEIR OWN BATTLE

Record labels aren't the only ones going to court. Jason Isbell, David Lowery, Guy Forsyth and Eduardo Calle filed a proposed class action against Suno.

Their case takes a different approach. Instead of centering primarily on ownership of sound-recording copyrights, the musicians allege that Suno improperly exploits artists' identities and enables outputs associated with particular musicians. The case raises right-of-publicity and related claims concerning an artist's name, likeness and identifying characteristics. Suno disputes the allegations and says its platform is designed to help people make original music.

That potentially opens another front in the AI debate. Owning a recording is one thing. Owning an artist's identity is another. And being able to generate something that sounds remarkably like a particular performer could create legal questions even when the resulting song isn't a direct copy of an existing recording.

EVEN THE LABELS' AI DEALS ARE BEING CHALLENGED

There's another twist. The American Federation of Musicians sued Universal Music Group and Warner Music Group, alleging that AI licensing arrangements involving musicians' recordings violated union agreements by allowing recordings to be used without appropriate compensation to the performers. Universal has disputed the allegations.

So now we have labels suing AI companies, artists suing AI companies, musicians suing labels over the labels' AI agreements, and portions of the industry negotiating licensing deals designed to make commercial AI music legitimate.

THE DJ HANGOUT TAKE

AI music isn't going away. The more important question is becoming who gets paid, who gives permission, and where the training material comes from.

For DJs and producers, this isn't some distant Silicon Valley legal argument. AI-generated vocals, stems, remixes, tracks and production tools are moving directly into the same ecosystem DJs already work in.

If courts ultimately decide that training on copyrighted recordings requires licensing—or that obtaining training material by circumventing technological protections creates separate liability—the economics of AI music could change dramatically. On the other hand, if AI companies prevail on major portions of their fair-use arguments, the technology could gain considerably more freedom to learn from existing music.

There may be a third outcome already emerging: AI music becomes a licensed business. Labels provide catalogs. Artists and musicians negotiate compensation and consent. AI companies get legitimate training material. Consumers get increasingly powerful creation tools.

That may eventually prove more commercially important than any individual lawsuit. For now, however, the lawsuits are multiplying faster than the settlements, and the battle over who controls the future of music is only getting started.