UMG, Sony Hit Suno With Second Copyright Infringement Action

The major labels are taking another swing at Suno.

Universal Music Group and Sony Music Entertainment filed a second copyright infringement action against the AI music company on Sept. 18, alleging that Suno copied tens of thousands of copyrighted sound recordings without permission to train the models powering its service.

Filed in the U.S. District Court for Massachusetts, the complaint identifies more than 60,000 such copyrighted sound recordings—though the label groups stress that this figure represents only a fraction of the works they allege Suno has infringed upon. The plaintiffs say that Suno’s concealment of its training data has prevented them from identifying the full scope of the alleged copying.

The RIAA brought the first major-label copyright suits against Suno and generative AI platform Udio on behalf of UMG, Sony and Warner Records in June 2024, alleging that they used copyrighted recordings to train their models without authorization. (Warner settled with Suno in November 2025.) Suno subsequently acknowledged obtaining audio from YouTube to train its AI and challenged UMG and Sony’s standing to pursue a related “stream ripping” claim.

The new complaint takes aim at Suno’s recently launched Version 6 models, alleging that they remain connected to the unlicensed training corpus even when the copyrighted recordings aren't used directly in the final round of training. According to the filing, Suno launched Version 6, Version 6 Wild and Version 6 Mini on Sept. 9 and acknowledged training Version 6 using “user interactions” with previous model iterations.

The labels argue that those interactions amount to another path by which copyrighted expression allegedly flows into training models. Per the legal filing, the earlier Suno models were themselves trained on the unlicensed recordings, meaning that what is derived from those outputs carry forward capabilities learned from the copyrighted material.

The complaint likewise alleges that Suno used “knowledge distillation” and related model-transfer techniques to develop Version 6, in which a new “student” model learns from a predecessor “teacher” model. The label groups characterize Suno's earlier models as teachers and argue that Version 6 therefore inherits capabilities derived from the copyrighted recordings.

“Training a ‘new’ model on the outputs of an infringing model does not eliminate the infringement; it launders it,” the complaint alleges. “V6 is not a fresh start; it is the fruit of the same poisoned tree.”

The filing further alleges that every Suno model from Version 4 through Version 6 has been tainted by the same original corpus of copied recordings. It names versions 4, 4.5, 4.5+, 5, 4.5-All and 5.5, arguing that even a model not directly trained on the plaintiffs’ recordings can allegedly benefit from Suno's unauthorized copies and data associated with user-generated outputs.

The labels also raises what they describe as the broader market impact of generative music, positing that Suno can generate enormous quantities of competing audio much faster than human creators can produce original recordings and citing a 2025 ruling in Kadrey v. Meta Platforms concerning market harm from generative AI training.

“A service that can produce ‘literally millions of secondary works’” in a fraction of the time and creativity required to make the originals, the filing argues, threatens to “dilute and ultimately overwhelm” the market for the recordings on which it was built.

Earlier this month, Jason Isbell led a group of musicians in filing a proposed class-action lawsuit against Suno, alleging that its technology allows users to generate songs that imitate artists simply by entering their names as prompts.