Warner Music Group struck a licensing deal with Suno. Sony Music and Universal Music Group are taking a different route. They now have two federal lawsuits in Boston over how the AI music platform built its models, including its newer v6 system.

How the fight started

In June 2024, record labels affiliated with Universal, Sony and Warner sued Suno in federal court in Massachusetts in a case coordinated by the Recording Industry Association of America. The labels alleged that Suno copied copyrighted sound recordings without permission to train a service that generates songs from prompts. They sought up to $150,000 per work for alleged willful infringement, the statutory ceiling they requested. Read the original complaint.

Warner took the deal

Warner and Suno announced their settlement and licensing partnership on November 25, 2025. The deal gives artists and songwriters the choice to opt in to uses of their names, images, likenesses, voices and compositions in new AI-generated music. Suno also acquired the concert-discovery service Songkick from Warner. Read Warner’s announcement.

On September 9, 2026, Suno introduced v6, saying it had developed the new generation with partners including Warner, BMG and Believe. The company said it would retire earlier models as v6 rolled out. Read Suno’s v6 announcement.

Why Sony and Universal filed a second case

In August 2026, the judge overseeing the original case declined to let the remaining labels add 61,026 recordings to that lawsuit. The original complaint had listed 560 examples, including Warner recordings. The judge separately allowed the labels to add a Digital Millennium Copyright Act claim alleging that Suno bypassed YouTube’s technical protections to obtain audio. Both rulings addressed what the labels could plead. Read the order on the additional recordings. Read the DMCA order.

On September 18, Universal- and Sony-affiliated labels filed a separate Boston lawsuit involving 60,202 sound recordings. They allege that v6 benefited from outputs and user feedback tied to older models they say were trained on unauthorized copies. Suno says v6 was trained using licensed partner content, community interactions and feedback, and knowledge its team had gained. Read the reporting on the second complaint.

In its October 20 answer, Suno acknowledged that it had retained some recordings used to train its AI models. The company said it preserved the material in line with a request from the labels before the original lawsuit. Which material went into v6 remains a separate factual dispute. Read Suno’s answer.

Germany ruled on six compositions. Boston is still pending

On July 31, a Munich court found in a separate case brought by German collecting society GEMA that Suno’s model had memorized six protected compositions. The court applied U.S. copyright law to training copies made in the United States and rejected Suno’s fair-use defense on those facts. It ordered Suno to stop specified uses and upheld claims for disclosure and damages involving several of the works. The ruling came from a first-instance German court and can be appealed. Read WIPO’s case summary. Read the European Commission’s case summary.

What to watch next

In a related New York subpoena dispute, Universal and Sony sought DistroKid data on how many tracks it delivered to streaming services were flagged as AI-generated and how many were attributed to Suno. A late-October joint filing, reported November 2, said the parties were working toward a resolution and moved DistroKid’s response deadline to November 6. Read the subpoena update.

In the original Boston case, the reported schedule set September 30, 2026, as the fact-discovery deadline and April 9, 2027, as the deadline for motions that could resolve some or all claims before trial. The Boston court has yet to decide whether the training at issue qualifies as fair use. See the reported court schedule.

Why it matters for producers

You can program a beat from scratch by writing code that synthesizes the sounds directly, with no samples or model-training data. The question in Boston is about a different step in music production: whether a company can copy other people’s recordings to train a model that generates music. The answer could shape how AI music companies license training material and how artists and rightsholders get paid for its use.