Universal Music Group and Sony Music Entertainment are suing Suno again, targeting the AI music company’s newly released v6 models and alleging that information derived from its earlier technology helped shape the new generation.
Filed September 18th in federal court in Massachusetts, the second lawsuit alleges infringement involving 60,202 sound recordings owned by the labels. Suno says v6 was built from scratch through a new system that includes licensed music. UMG and Sony argue the break from Suno’s earlier models is not that clean.
Their new complaint focuses partly on whether outputs and user preference signals generated by older Suno models carried knowledge into v6. In other words, the fight is no longer limited to asking which recordings were directly placed into an AI training dataset.
UMG and Sony describe the new models as the “fruit of the same poisoned tree,” Music Business Worldwide notes. Those allegations have not been proven, and Suno has disputed the record companies’ broader copyright theories.
Suno v6 Was Supposed to Change the Equation
Suno launched v6 on September 9th as a major reset for its music-generation technology.
Unlike the earlier models at the center of existing copyright litigation, the new generation was developed alongside music companies that have reached agreements with Suno. Warner Music Group, BMG and Believe are among the rightsholders participating in its newer licensing framework.
TechCrunch detailed the transition when v6 launched, including Suno’s position that the model did not rely on the same training dataset behind its previous versions.
Three versions arrived as part of the reset. Suno introduced its standard v6 model alongside v6-wild, a more experimental option for paying users, and a free version with additional restrictions.
For a company already defending itself against major copyright claims, the distinction between old and new technology carries obvious legal weight. Warner settled its lawsuit against Suno and entered a licensing partnership, while BMG reached an agreement before v6 arrived.
Universal and Sony chose a different path. Both remain plaintiffs in the original litigation over Suno’s earlier training practices, and the second lawsuit challenges whether v6 can truly be treated as independent from what came before it.
User Preference Signals Are Now at the Center
One of the most important parts of the new case involves something that is not itself a recording: what Suno learns from the choices its users make.
Suno has discussed using preference signals to improve its technology. Imagine the system generates Song A and Song B, then asks listeners which version they prefer. Repeated across large numbers of interactions, those choices create information about what people tend to like.
CEO Mikey Shulman discussed that approach in an interview with Billboard, describing user preference data as part of the company’s effort to improve future music generation.
From Suno’s perspective, that kind of information is different from putting a copyrighted Sony or Universal recording directly into a training dataset. A user choosing Song A over Song B is supplying a preference signal, not uploading another label-owned master recording.
UMG and Sony want the court to look one step further back.
Their complaint alleges that v6 was developed using “outputs of, and preference signals derived from” earlier Suno models that the labels contend were themselves trained through unauthorized copying. The Hollywood Reporter highlights that allegation as a central part of the new case.
Consider the Song A versus Song B example again. If both recordings were generated by an older model, a listener’s preference between them still tells Suno something about which qualities produced by that model people respond to.
UMG and Sony argue that information may represent another way the benefits of earlier training can carry forward even when the original recordings are no longer directly included in the newest training dataset.
Whether copyright law treats those preference signals the way the labels want it to is unresolved. Their theory remains an allegation, but it pushes the Suno litigation into territory more complicated than simply proving whether a copyrighted song was copied during training.
Knowledge Distillation Opens a New Copyright Question
The allegations also touch on a broader machine-learning concept commonly known as knowledge distillation.
Broadly speaking, a newer AI model can sometimes learn from the behavior or outputs of another model instead of learning exclusively from the original material used to train that earlier system.
Think of a student learning from a teacher who already absorbed thousands of recordings. The student may never hear those exact records, but some of what the teacher learned can still influence the lesson.
AI models do not learn like human musicians, and the analogy should not be taken literally. It does, however, illustrate the copyright question UMG and Sony are raising: if information extracted from an older model helps train a new one, how separate are the two generations really?
That question matters because Suno’s legal position around v6 rests partly on the idea of a fresh start. Newly licensed music and different training practices potentially create a clearer path forward than continuing to rely directly on material disputed in the original lawsuits.
If v6 learned only from newly licensed recordings and unrelated data, the separation is easier to describe. If outputs, rankings or other information from older models materially helped teach it, UMG and Sony argue that the lineage deserves legal scrutiny.
No court has established that such model-derived information automatically carries copyright liability. Determining whether the labels can successfully connect those techniques to infringement is part of what makes the second lawsuit potentially important beyond Suno itself.
AI developers increasingly use synthetic data, model outputs and machine-generated information when building newer systems. Music copyright disputes may therefore have to address not only what went into the first model, but what one generation of technology can legally teach the next.
The Complaint Covers 60,202 Recordings
Scale gives the new lawsuit another striking dimension.
UMG and Sony identify 60,202 sound recordings in their complaint, MBW states. The labels accuse Suno of copying protected recordings without authorization while developing its technology.
U.S. copyright law allows statutory damages of up to $150,000 per work for willful infringement. Multiplying that maximum by every recording identified would create a theoretical ceiling above $9 billion.
That number should not be mistaken for a prediction of what the labels will receive or what a court would award. Liability has not been established, statutory damages depend on multiple legal findings, and maximum damages are not automatically imposed for every work identified in a complaint.
Another part of the dispute involves music obtained from YouTube.
Earlier in September, Suno acknowledged in the original litigation that audio data had been obtained using YT-DLP for training purposes. Suno’s court filing states that audio was obtained using the tool, while the company disputes whether UMG and Sony have standing to pursue the related circumvention claim.
The labels are seeking damages for alleged copyright infringement as well as claims connected to the alleged circumvention of YouTube’s technological protections. Suno continues to dispute the broader infringement allegations and has argued in the existing litigation that its training practices are protected by fair use.
Warner Took a Different Route With Suno
Warner Music Group’s absence from the second lawsuit tells its own music-industry story.
Warner originally joined UMG and Sony in litigation against Suno but settled its dispute in 2025 and entered a licensing partnership with the company. BMG and Believe have also struck agreements connected to Suno’s newer generation of products.
Major rightsholders are therefore pursuing two strategies at once: licensing AI music systems under negotiated terms or continuing to challenge how those systems were built.
Suno chief product officer Jack Brody has described the newer phase as an opt-in framework in which participating rightsholders can share revenue and artists can eventually decide whether to participate in AI experiences connected to their work. Axios outlined those arrangements when v6 launched.
That leaves Suno in an unusual position. One part of the established record business is supplying licensed material and participating economically in its newest technology, while two of the world’s largest music companies argue that the same generation still benefits from earlier unauthorized copying.
The divide is becoming less about whether generative music should exist at all and more about the terms under which it can be built: what music can train a model, who grants permission and how artists and rightsholders participate financially.
This Is Not the DistroKid Lawsuit
The timing makes UMG’s two new lawsuits easy to confuse, but they target different stages of digital music.
UMG sued DistroKid on September 15th. Three days later, it joined Sony in filing the new Suno case. Both involve AI-generated music and copyright, but the central allegations are different.
- DistroKid: UMG alleges that a distributor delivered unauthorized, altered and AI-generated recordings into commercial streaming services.
- Suno: UMG and Sony allege that an AI company copied protected recordings and that information derived from earlier models continues to benefit its v6 technology.
The simplest distinction is creation versus distribution.
Suno generates music. DistroKid provides infrastructure for creators and artists to deliver recordings to services such as Spotify and Apple Music.
UMG’s DistroKid lawsuit focuses on responsibility for material moving through a distributor and onto streaming platforms. The Suno litigation reaches further upstream into how generative systems acquire the ability to make music in the first place.
Both cases could affect the emerging AI music economy, but they test different parts of it.
Suno Faces More Than One Copyright Fight
Separate litigation is also testing whether generative music can cross another line: artist identity.
Earlier in September, musicians including Jason Isbell, David Lowery, Guy Forsyth and Eduardo Calle filed a proposed class action alleging that Suno can generate music evoking specific performers even when artist names are restricted in prompts.
Those musicians argue that the system commercially exploits their identities without permission. Suno denies the allegations and says its technology is designed to create original music rather than impersonate artists, Reuters details.
That case raises a different legal theory from UMG and Sony’s copyright claims. Combined with the original major-label lawsuit and the new v6 complaint, however, it shows how many parts of Suno’s technology are now being tested in court.
The September 18th lawsuit pushes the fight into particularly new territory. UMG and Sony are no longer challenging only the older Suno models built before the company began signing major licensing deals. They are asking whether the generation that followed can legally be separated from what those earlier systems learned.
Suno has not yet filed its formal response to the new complaint.