Sony Music and Universal Music Group have filed another copyright lawsuit against AI music company Suno, arguing that its version 6 model still benefits from recordings allegedly copied to train earlier systems. The labels describe the process as “model laundering”: using outputs from an older model to teach a nominally new one.
Suno said when version 6 launched that it had been trained from the ground up on a new dataset that included user data. The company later confirmed that the material included creations made by users, but did not provide further detail. Sony's complaint alleges that Suno also used distillation, a technique in which one model learns to reproduce behavior or outputs from another model.
The labels' argument is that a fresh model is not independent if its training targets encode patterns learned from unauthorized copies. Under that theory, version 6 would remain connected to recordings allegedly taken from YouTube and other sources even if those files were not placed directly in its own training set. These are allegations in a complaint, not a court finding.
Sony and Universal have not joined licensing agreements that Suno reached with other music companies. The case could help define whether training on synthetic outputs changes the copyright analysis or simply transfers disputed material through another generation of models.