Sony Music Entertainment has opened a second front in its war with the artificial-intelligence music startup Udio, filing a fresh federal lawsuit that accuses the company of copying more than 30,000 of its recordings to train the models that generate songs on demand. The complaint, lodged in the U.S. District Court for the Southern District of New York, puts a precise and startling number on the alleged infringement: 30,117 tracks, spanning catalog as varied as Beyoncé, Harry Styles and Elvis Presley.
That figure did not come from guesswork. According to the filing, it surfaced through discovery in Sony's original case against Udio — the legal process turned up the recordings sitting inside the startup's training datasets. In other words, Sony says it now has a receipt.
Why Sony had to sue twice
The obvious question is why a company already suing Udio would need to sue it again. The answer is procedural, and it matters. When Sony tried to fold these 30,117 newly identified recordings into its existing complaint, the presiding judge, Alvin K. Hellerstein, denied the request. Rather than let the tracks go unaddressed, Sony filed an entirely separate lawsuit to cover them — a move that protects the copyrights now while the original case grinds on.
The result is two parallel actions against the same defendant, and a much larger universe of allegedly infringed works than the first suit contained.
The money is staggering on paper
Sony is demanding a jury trial and seeking statutory damages of up to $150,000 for each infringed work. Do the arithmetic on 30,117 recordings and the theoretical exposure climbs into the billions of dollars — north of $4.5 billion at the maximum. Statutory-damage ceilings are rarely awarded in full, and the real figure, if any, would be set by a jury or a settlement. But the number establishes the leverage Sony is bringing to the table, and it signals that the label has no interest in a quiet resolution.
A fight the rest of the industry already stepped back from
This is where Sony's position becomes conspicuous. The music business first moved against AI generators in June 2024, when Sony, Universal Music Group and Warner Music Group jointly sued both Udio and its rival Suno, arguing the companies had built their products on mountains of copyrighted recordings without permission.
Since then, the united front has fractured. Universal and Warner reached licensing agreements with Udio, converting adversaries into business partners and effectively deciding that a deal beat a courtroom. Sony did not. Its decision to escalate — filing a second suit rather than settling into a license — leaves it as the major label still willing to litigate the underlying question rather than negotiate around it.
The question a jury may have to answer
Udio has not disputed that copyrighted music is in its training pipeline. In an earlier response to Sony's original suit, the company acknowledged training its models on copyrighted songs uploaded to YouTube, but argued that doing so amounts to fair use — the legal doctrine that permits limited use of protected work without a license under certain conditions.
That single argument is the heart of the entire dispute, and not just for these two companies. Whether feeding copyrighted recordings into an AI system to teach it how to make new music counts as fair use or as mass infringement is one of the defining unsettled questions of the generative-AI era. Courts have only begun to grapple with it, and a jury verdict in a case involving 30,117 identified recordings and a household-name plaintiff would carry weight far beyond Sony and Udio.
For now, the immediate stakes are concrete. Sony has drawn a hard line where two of its biggest rivals chose to make peace, and it is asking a court to put a price — potentially an enormous one — on how AI music gets built. The outcome will help decide whether the next generation of these tools is trained on licensed catalogs by agreement, or fought over one recording at a time.
