Sony Sues Udio Over More Than 30,000 Songs In AI Music Fight

Sony Music has filed a new lawsuit against Udio, escalating one of the music industry’s most important legal fights with generative AI. The new case names more than 30,000 songs that Sony says were infringed by the AI music generator, turning what was already a major copyright battle into a much larger test of how music training data will be treated in court.

The Verge reported that Sony’s latest complaint was filed in New York and follows an earlier 2024 lawsuit against Udio. Sony says it identified the additional recordings after gaining access to Udio training data during discovery and using audio fingerprinting to compare material.

The list reportedly includes major works from artists including Elvis Presley, Beyonce, Britney Spears, Johnny Cash and Harry Styles. Sony is seeking to block further alleged infringement and is asking for statutory damages that can reach up to $150,000 per infringed work under U.S. copyright law.

Sony had previously tried to add more than 30,000 recordings to the existing Udio lawsuit, but a New York federal judge refused to expand that case so late in discovery. Music Business Worldwide reported that the original case remained limited to 333 works after the court rejected Sony’s motion.

The new lawsuit appears to be Sony’s answer to that procedural problem. Instead of trying to force a much larger catalogue into the older case, Sony is putting the additional songs into a fresh complaint. That means the court will still have to confront the larger question: whether Udio copied protected recordings to train a model that can generate music on demand.

Udio has argued in the past that training on existing music should be protected by fair use. The music industry disagrees. Labels and publishers say AI music systems cannot build commercial products from copyrighted recordings without permission and then compete with the artists and rights holders whose catalogues made the systems possible.

The Udio situation is especially interesting because the industry is not taking one single path. Universal Music Group and Warner Music Group have already settled with Udio and moved into licensing relationships. Sony, however, remains in court.

That split tells us where the market is heading. Some rights holders want to litigate hard enough to define boundaries. Others want licensing deals that turn AI music into a controlled business line. Both strategies can exist at the same time because lawsuits create pressure, while licences create revenue.

A similar pattern is playing out across AI copyright. The Anthropic author settlement shows that legal risk can become very expensive. The AI music cases may do the same for sound recordings, compositions, artist likeness and style imitation.

Sony’s use of audio fingerprinting is important because music copyright disputes often depend on proof. It is not enough to say an AI model sounds influenced by a catalogue. Plaintiffs need evidence that recordings were copied, ingested or reproduced in a way the law recognises.

Audio fingerprinting can help rights holders compare training data, outputs or internal datasets against known recordings. If courts accept that evidence, labels will have a stronger route for identifying alleged infringement at scale. That could change how AI music companies store logs, disclose training sources and defend their models.

The legal fight will also test what it means for a model to “learn” from music. Is training closer to a human musician listening widely, or closer to mass copying for a commercial machine? That question sounds philosophical, but the answer will decide billions of dollars in licensing value.

For artists, the fear is not only that songs were used without permission. It is that AI tools could flood streaming platforms, advertising, games and social media with music that competes against human creators while drawing from their work. Even if outputs are not exact copies, the market impact could be real.

There is also the issue of style. A model trained on thousands of recordings can produce music that feels close to a genre, era or artist signature. Copyright law traditionally protects specific works more strongly than broad style, but AI is blurring that line because style can now be generated at industrial scale.

This is why the Sony-Udio case is bigger than one company. It is about whether the music industry can force AI music platforms into a permission-based model before AI-generated songs become normal content on streaming services and social platforms.

The lawsuit arrives as AI copyright disputes are becoming more concrete. Book authors are settling with AI labs. Publishers are suing model makers. Music labels are forcing AI audio companies toward licences. The next phase will likely be less about whether AI should exist and more about who gets paid when copyrighted culture is used to build it.

For Udio, the case could determine whether it remains a disruptive AI startup or becomes another licensed technology provider inside the music business. For Sony, it is a chance to defend catalogue value before AI music becomes too embedded to unwind.

For the rest of the tech industry, the warning is simple: training data is no longer invisible. Rights holders are finding ways to trace it, price it and sue over it.