Suno Says User Creations Helped Train Its Music Model Amid New Copyright Suit

Suno says music created through its platform helped train its latest AI music model. That disclosure puts user creations at the center of a new copyright fight: Sony Music Entertainment and Universal Music Group allege the model carries forward material from earlier systems they say used their recordings without permission.

The labels filed their second copyright lawsuit against Suno in Boston federal court on September 18. They allege it copied 60,202 sound recordings they control without authorization to develop its AI music technology. Suno’s September 22 response supplied a sharper description of its V6 training: licensed content from music partners, plus creations and preference signals from interactions with users.

The filing came nine days after Suno launched V6, a model it described as developed with Warner Music Group, BMG and Believe.

What Suno says went into V6

Suno draws a distinction between audio uploaded by users and music created through its service. It previously said the user data used to train V6 did not include uploaded audio files. Its later statement identifies user creations and preference signals as training inputs, alongside licensed material. The company has not specified what share came from user-generated music or identified the individual recordings used.

Suno’s terms of service, effective September 3, give it a broad, nonexclusive license to use submitted and generated content for purposes that include improving its AI models. That permission does not automatically transfer ownership of users’ original submissions to Suno. The terms address Suno’s use of content on its platform; the labels’ suit raises a different question about the training history of the models that produced some of it.

The alleged link between old and new

Sony and Universal argue that licensing some training material for V6 would not necessarily break its connection to earlier models. Their complaint alleges V6 was trained partly on music those models produced and that Suno used techniques to transfer capabilities developed through earlier training. In their account, a newer model can learn from an older model’s output even without directly receiving the recordings at issue.

The labels call this alleged process “model laundering.” The phrase is their argument, not a court finding. It describes their concern that a second model could inherit characteristics learned from copyrighted recordings through music generated by a first model. Suno calls the labels’ claims fundamentally flawed on both the facts and the law. No final judicial determination has been made in the new case.