Sony and UMG sue Suno over v6 model laundering

Major record labels Sony and Universal Music Group have launched a new legal battle against generative audio startup Suno. The lawsuit targets Suno's latest v6 model, accusing the company of engaging in "model laundering." According to the plaintiffs, Suno trained its newest system on user outputs and data generated by its older models, which the labels argue were built using unlicensed, copyrighted music scraped from YouTube and other online platforms.

When Suno released v6, the company's Jack Brody asserted that the model was built from scratch using a fresh dataset. Suno spokesperson Rachel Racusen later clarified that this training material included user interactions, creations, and community preference signals, according to statements updated on September 25th, 2026. However, the record labels argue that this approach does not clean the data. Sony alleges that Suno utilized a process called distillation, effectively training the v6 model to mimic the outputs of its older "teacher" models. The plaintiffs argue that because those original models relied on unauthorized copies, any subsequent model trained on their outputs remains legally compromised.

For AI developers and music creators, this lawsuit highlights the growing legal risks surrounding model distillation and synthetic training data. If the courts agree with the labels, developers will not be able to bypass copyright liabilities simply by training new models on the outputs of older, disputed systems. Practitioners may need to implement strict provenance tracking to prove that their training pipelines are entirely free of "tainted" synthetic data, potentially raising the cost and complexity of developing next-generation generative media tools.