Universal Music Group and Sony Music Entertainment have filed a second federal copyright lawsuit against AI music generator Suno, accusing the company of copying tens of thousands of their recordings and arguing that its new “v6” models remain tainted by earlier unlicensed training.
The joint complaint was filed on Friday, 18 September 2026, in Boston federal court. It names 60,202 sound recordings that the labels say were copied without a licence and used to build the models behind Suno’s generation service. Under US law, willful infringement can carry statutory damages of up to $150,000 per work — a theoretical ceiling of just over $9 billion if applied to every title in the new complaint. The labels also seek up to $2,500 per recording for alleged circumvention of YouTube’s anti-download protections, an injunction blocking further use of their catalogues, and a jury trial.
Why a second case, and why now
UMG, Sony and Warner Music Group first sued Suno in June 2024. Warner later settled and licensed its catalogue to the company in November 2025. Universal and Sony stayed in that original case, which still covers a smaller illustrative set of works.
In August 2026, Judge F. Dennis Saylor IV refused a request to expand that first lawsuit with more than 61,000 additional recordings, saying the extra titles would delay a fair-use decision and noting that the majors could file a separate suit. Friday’s complaint is that separate action. The labels say audio fingerprinting during discovery in the first case helped them identify their works inside Suno’s training data — and that even 60,202 titles still represent only a fraction of what they believe was copied.
The “poisoned tree” attack on v6
The new filing’s centrepiece is Suno’s v6 suite, launched on 9 September 2026 and promoted as built in partnership with Warner, BMG and Believe. Suno’s chief product officer, Jack Brody, told Music Business Worldwide at launch that v6 was “trained entirely from scratch” on data that “doesn’t include data from Universal or Sony.”
Universal and Sony reject that clean break. They allege v6 was trained on “user interactions” with earlier models — outputs and preference signals generated by systems that, they say, were themselves trained on unlicensed catalogues. Training a new model on those outputs, the complaint argues, “does not eliminate the infringement; it launders it.” Their phrase for the result: v6 is “the fruit of the same poisoned tree.”
The complaint also alleges knowledge distillation, in which a “student” model learns to reproduce a “teacher” model’s behaviour. The teachers, Universal and Sony say, are prior Suno versions (including v4 through v5.5) whose musical capabilities were derived from copied recordings.
Fair use, licensing markets, and market harm
Suno’s core defence in the first case is that training on copyrighted recordings is “quintessential fair use.” One factor courts weigh is whether unlicensed use harms an existing market. Universal and Sony argue Suno’s own recent deals undercut that defence: settlements and partnerships with Warner, BMG and Believe, they write, show a functioning market for licensed AI training data in which Suno is now a paying participant.
On harm to human repertoire, the complaint cites streaming dilution. It points to Deezer’s July 2026 disclosure that more than half of new tracks arriving on its platform each day — around 90,000 — were AI-generated. Machine-generated tracks, the labels say, compete for placement, discovery, listener attention and royalty pools.
Suno, in a statement to Billboard, called the new claims “fundamentally flawed on both the facts and the law,” said v6 was trained on licensed partner content plus community interactions and team learnings, and rejected the idea that its product is designed to imitate specific artists’ work.
What this means for creators and platforms
The dual track — an original fair-use case plus a second suit aimed at v6 and a much larger work list — keeps the industry’s central AI-copyright question in front of US courts while commercial licensing experiments continue. Warner’s settlement path and Universal–Sony’s litigation path now coexist as rival strategies for the same technology.
For artists and rights holders, the dispute is about control as much as damages: whether generative tools can keep building on catalogues without consent, and whether “synthetic” retraining can reset that history. For Suno and peer startups, the risk is not only statutory exposure but product strategy — how far partnerships can rehabilitate models whose earlier training remains contested.
Fact discovery in the first Universal–Sony case closes on 30 September 2026. The second complaint will move on its own docket. Neither filing alone settles the fair-use question that still hangs over generative AI music; together they raise the cost of leaving that question unanswered.