Universal and Sony sue Suno again, this time over its brand-new v6 model

The fight over AI-generated music opened a new front on 18 September. Universal Music Group and Sony Music Entertainment filed a second lawsuit in Boston federal court against the AI music company Suno. It accuses Suno of copying 60,202 of the labels' sound recordings without a licence to build its models. These are allegations, and Suno has argued in the earlier case that training on copyrighted recordings is fair use. The new complaint is still worth reading closely, because it tests an idea the whole AI industry cares about.

The "poisoned tree" argument

Suno launched a new model suite, v6, on 9 September. It says v6 was built with Warner Music Group, BMG and Believe. Its chief product officer told Music Business Worldwide that v6 was trained from scratch, on data that doesn't include Universal or Sony material.

The labels say that doesn't matter. Their complaint argues that v6 learned from "user interactions" and preference data. In other words, it learned from people choosing between tracks generated by Suno's earlier models, which the labels say were trained on their recordings. They also allege Suno used knowledge distillation, where a new "student" model learns to imitate an older "teacher". Training on the outputs of an infringing model, they argue, launders the infringement rather than ending it. The complaint's memorable phrase: v6 is "the fruit of the same poisoned tree".

Why a second case?

Universal and Sony first sued Suno in June 2024 over 560 works. In August 2026 they asked to add more than 61,000 recordings to that case. The judge declined, saying it would delay a case he wants decided on fair use, and noting the labels could sue separately. So they did. The labels say they identified their recordings in Suno's training data with the help of the audio-fingerprinting company Audible Magic.

The market-harm angle

A key part of any fair-use defence is whether the new use harms the market for the original. Here the labels point to Suno's own licensing deals as proof that a market for training data exists. They also cite the French streaming service Deezer, which said in July that more than half of the new tracks arriving each day were AI-generated.

What it means for creators

Nothing is decided yet, and these are claims, not findings. But the case could shape how every generative tool, for music, images or text, has to account for its training data. Even a model that's "new" may carry its history with it.

For artists and makers, the practical lesson is clear. Provenance matters more every month. Work that's clearly yours, made by hand and signed with your own style, is becoming its own kind of value.