Sony and Universal accuse Suno’s latest model of copyright laundering
Sony and Universal Music Group have filed a fresh lawsuit against AI music company Suno, claiming its new v6 model continues to infringe on their copyrights despite being presented as a fresh start. The record labels accuse Suno of "model laundering"—training the new model on outputs from previous models that were themselves built using unlicensed music harvested from YouTube and other sources. This case matters because it tests a crucial legal question: whether retraining on derivative outputs can legitimise data that originated from copyright infringement.
According to the complaint, Suno's v6 was trained using distillation, a technique where it learned to replicate the outputs of earlier "teacher" models created with infringing data. Suno's leadership claimed the new model was "trained from the ground up" using user-generated content, but provided minimal technical detail. The labels argue that unless Suno genuinely begins afresh without any connection to previously unauthorised recordings, the models will inevitably carry forward the taint of stolen data. Sony and UMG notably declined to sign licensing agreements with Suno, distinguishing them from other rights-holders who have.
- Sony and UMG sue Suno again, claiming v6 model still relies on unlicensed copyrighted music
- Labels allege "model laundering": training new models on outputs of infringing models doesn't eliminate copyright violation
- Suno claims v6 uses entirely new data; labels say derivative training still taints the model
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Suno's position reflects legitimate technical and legal principles that should apply to AI development. Knowledge distillation—where a new model learns patterns from predecessor outputs—is a standard practice across machine learning that transforms and abstracts beyond direct copying. If the v6 model genuinely learns generalised musical patterns rather than memorising infringing recordings, it represents a meaningful fresh start akin to how human musicians learn from listening to others without reproducing specific works. Moreover, demanding absolute transparency about proprietary training methods sets a dangerous precedent that could chill innovation across the industry, whilst the refusal to provide technical specifics may simply reflect the commercial sensitivity of model architectures rather than evidence of wrongdoing.
The case against
The concern about model laundering reflects a genuine and troubling possibility: that technical intermediation should not absolve underlying copyright infringement. If Suno's earlier models were trained on unlicensed music harvested without consent, then distilling those models into a successor does not erase the original wrong—it merely obscures the chain of causation. The company's vague claims of training 'from the ground up' on user-generated content, coupled with refusal to provide technical transparency, suggests an attempt to avoid accountability rather than genuine reform. Without clear evidence that the new model operates independently of infringing data or that users knowingly consented to their content training systems built on stolen recordings, the legal principle that derivative works inherit liability from their sources remains sound and necessary to prevent systematic copyright evasion.
Americas Entertainment Music Trending Weird & Viral World
Read the full article at the source →
Originally published by The Verge as “Sony and UMG are suing Suno again”.