Sony and Universal Music Group have filed a new lawsuit against Suno, claiming its latest AI music generation model, v6, infringes on copyrights by building on outputs derived from unlicensed music. The labels assert the model remains tainted by prior unauthorized use of protected works.
- Sony and UMG accuse Suno's AI model of 'model laundering' copyrighted music.
- Suno claims v6 was retrained with licensed content and user-generated data.
- The case raises questions about copyright risks in AI-generated music.
What happened
Sony and Universal Music Group have launched a renewed lawsuit against Suno regarding its AI-driven music generation platform. The suit centers on Suno's version 6 (v6) model, which the labels say is still indirectly trained on copyrighted music ripped from sources like YouTube, despite claims of retraining. The labels argue that v6's training includes user outputs from earlier models built with unlicensed data, which constitutes continued infringement.
Sony and UMG have not entered into licensing agreements with Suno, in contrast to other industry entities that have struck deals allowing AI use of their music catalogs. The complaint specifically accuses Suno of 'model laundering,' where one AI model leverages outputs generated by a previous, infringing model, thus perpetuating unauthorized use. Suno representatives assert that v6 incorporates licensed content as well as community-generated interactions but have not detailed the extent or nature of this material.
Why it matters
This lawsuit highlights ongoing tensions between music rights holders and AI companies developing generative tools using massive datasets. The music industry is concerned that unauthorized use of copyrighted works to train AI models could dilute artists' control and revenue streams. Sony and UMG’s actions demonstrate that major labels remain willing to pursue litigation to protect their intellectual property in the evolving digital and AI landscape.
For AI developers, the case underscores the complex legal and ethical challenges around sourcing training data. The concept of retraining or 'distillation,' where a new model is built to emulate outputs of a prior model trained on unlicensed content, may not shield companies from infringement claims. This dispute signals potential risks for other startups and platforms attempting to build on or refine existing AI models within copyrighted creative domains.
What to watch next
Industry stakeholders will closely monitor how courts interpret copyright law in relation to successive AI model training and derivative outputs. The resolution may set precedents on whether retraining AI using outputs from prior infringing models constitutes new infringement or legitimate use. This could influence how AI music companies approach licensing and dataset curation going forward.
Suno’s next moves will also be significant, including whether it can clarify the nature of its training data, negotiate licenses with major labels, or adjust its technology to avoid infringing activity. The broader AI music startup ecosystem will be watching this case to understand acceptable practices and the legal boundaries for developing innovative music generation tools.