Suno Copyright Lawsuit Puts Its Licensed AI Reset on Trial

Suno faces a second copyright lawsuit covering 60,202 recordings, with Universal Music Group and Sony Music challenging its attempt to restart AI music on licensed data. The Suno copyright lawsuit carries a theoretical statutory ceiling above $9 billion, although no court has awarded that amount.

The filing arrived nine days after Suno introduced v6, a model family developed through partnerships with Warner Music Group, BMG, and Believe. Suno presented v6 as a new beginning built around licensed music, artist participation, and revenue sharing.

Universal and Sony reject that clean-break narrative. They allege that knowledge inherited from earlier systems carried protected expression into v6, even if its direct training set excluded their catalogs.

That claim turns the case into more than a dispute over copied files. The central question is whether an AI developer can separate a new model from the disputed data, outputs, preferences, and technical learning behind its predecessors.

The New Suno Copyright Lawsuit Covers 60,202 Recordings

Universal and Sony have expanded their case from a representative copyright dispute into a catalog-scale challenge to Suno’s training history.

Twelve label entities filed the new complaint against Suno on September 18, 2026, in the US District Court for the District of Massachusetts. The federal case docket identifies UMG Recordings, Capitol Records, Sony Music Entertainment, and nine affiliated labels as plaintiffs.

The complaint asserts claims involving 60,202 sound recordings. The labels allege that Suno copied those recordings without permission while developing the models behind its music-generation service.

That number explains the $9 billion headline, but it requires careful interpretation. US copyright law allows statutory damages of up to $150,000 per work when a court finds willful infringement.

Multiplying that maximum by 60,202 produces a theoretical ceiling slightly above $9 billion. It is not a demand that a judge has approved, a settlement valuation, or a prediction of the final result.

Courts can award much less than the statutory maximum. The parties could also settle, narrow the asserted works, or resolve controlling legal questions before damages receive a full trial.

The labels additionally allege violations tied to circumventing YouTube’s technical restrictions. Their filing claims Suno used stream-ripping methods to obtain audio for model development.

A separate damages provision permits up to $2,500 for certain circumvention violations. Applying that maximum once to every asserted recording would add roughly $150 million, according to the labels’ theory.

Suno has acknowledged obtaining online audio with yt-dlp, a tool that can download media from supported websites. It disputes the legal significance of that conduct and challenges the labels’ broader infringement arguments.

The new action follows a procedural decision in the original 2024 case. Universal, Sony, and other plaintiffs initially identified 560 recordings as examples of works allegedly copied by Suno.

During discovery, the labels sought to add 61,026 works to that existing case. Judge F. Dennis Saylor IV rejected the amendment without preventing the labels from pursuing those claims separately.

His August court order said adding tens of thousands of works would create substantial complexity and delay. The original litigation was already approaching the end of fact discovery.

The judge also noted that Suno’s fair-use defense could resolve the predominant legal issue. Fair use permits some unlicensed uses of copyrighted material after courts weigh purpose, character, amount, and market harm.

A separate case therefore offered a way to preserve the additional claims without derailing the older lawsuit. The new complaint asserts 60,202 recordings rather than the full 61,026 previously proposed.

Universal and Sony say audio fingerprinting helped them identify their recordings within materials examined during discovery. Audio fingerprinting compares distinctive acoustic signatures rather than relying only on song titles or descriptive metadata.

Suno has not conceded that every listed recording was copied unlawfully. Nor has a court ruled that the fingerprinting process proves infringement across the entire catalog.

The case consequently begins with a large factual dispute. The labels must connect specific protected recordings to legally actionable copying, while Suno can challenge both their evidence and legal theory.

Still, the scale changes the negotiating environment. A dispute involving hundreds of examples can test a principle, while one involving 60,202 works can threaten an entire development strategy.

That difference creates the article’s core tension. Suno has shifted toward licensing, yet Universal and Sony argue that its new model remains inseparable from its unlicensed past.

Suno v6 Was Supposed to Mark a Licensed Reset

Suno designed v6 to show that generative music could move from litigation toward licensing, but the new filing attacks the reset itself.

Suno released v6 on September 9, 2026. The family includes a main v6 model, an experimental v6-wild option, and a faster v6-mini version for free users.

The company says v6 can follow detailed instructions, accept several input formats, and support precise edits to portions of a song. Those product claims have not been independently verified across every use case.

More important for this lawsuit, Suno says it built the models through agreements with established music companies. Warner Music Group, BMG, and Believe supplied the commercial foundation for that transition.

Suno’s chief product officer, Jack Brody, said the system was built from the ground up. He also said its training data did not include music controlled by Universal or Sony.

According to an account of the v6 training approach, Suno used licensed Warner recordings and Suno user data. The company also incorporated technical and preference learning from earlier systems.

That final point now carries unusual legal weight. A preference signal records how people rank or respond to model outputs, helping developers steer later systems toward desirable results.

Suno describes those signals as user feedback and general product learning. Universal and Sony characterize them as information generated through models allegedly trained on unauthorized copies.

The distinction matters because Suno is not merely promising future compliance. It is arguing that v6 belongs to a different technical and commercial era from its earlier models.

The company plans to retire those earlier systems. It has also described forthcoming opt-in experiences that would let participating artists authorize certain generative uses and receive compensation.

Those plans reflect a broader change in AI music. Labels increasingly treat licensing as an available business model, rather than opposing every generative product as inherently unacceptable.

Warner settled its dispute with Suno and became a partner. BMG and Believe also reached agreements that supported the new model generation.

Universal and Sony chose continued litigation. Their position creates a divided industry response to the same developer, with some rightsholders licensing Suno while others seek liability findings.

The divide pressures Suno in two directions. It must satisfy its partners with a commercially useful model while convincing a court that earlier development does not contaminate v6.

Users add another source of pressure. Retiring old models can disrupt established workflows, especially when creators prefer their sound, controls, or output consistency.

Reports following the v6 launch described complaints about output quality and tighter restrictions. Individual reactions do not prove broad rejection, but they expose the cost of replacing a popular model under legal pressure.

Suno cannot solve that tension by keeping every old system online indefinitely. Those systems sit at the center of the infringement allegations that its licensing strategy was meant to address.

The company also cannot assume that a licensed model will automatically match the behavior users liked. Training-data changes, safety controls, and new commercial rules can alter performance in ways that users notice immediately.

That makes v6 both a legal response and a product gamble. Suno needs its new system to support licensed growth without appearing dependent on the disputed models it replaces.

The timing of the lawsuit sharpens that gamble. Universal and Sony filed only days after Suno promoted v6 as evidence that collaboration with the music business was possible.

Their action effectively says that licensing future inputs does not erase responsibility for earlier copying. It also says that technical inheritance can keep that responsibility alive inside later products.

This is why the Suno copyright lawsuit reaches beyond one training dataset. It tests whether a company can establish a defensible boundary between generations of an evolving AI service.

Universal and Sony Say Technical Inheritance Carries Liability Forward

The labels’ most consequential argument is that a fresh dataset cannot cleanse a model built with knowledge derived from disputed predecessors.

Universal and Sony describe v6 as the product of the same allegedly infringing development chain. Their complaint points to synthetic training data, user preferences, and knowledge distillation.

Synthetic training data consists of material generated by another model rather than collected directly from original human-created sources. Developers can use it to expand datasets, teach behaviors, or transfer capabilities.

Knowledge distillation trains a smaller or newer student model to reproduce useful behavior from a teacher model. It can transfer capabilities without copying every component of the teacher’s architecture.

Neither technique is automatically unlawful. The unresolved question is whether their use can transmit legally protected expression derived from unauthorized source material.

The labels say earlier Suno systems learned from their recordings. They then argue that outputs and behavioral knowledge from those systems carried expressive features into v6.

Their complaint calls this process laundering rather than a fresh start. A detailed lawsuit account reports that the labels describe v6 as the “fruit of the same poisoned tree.”

That phrase summarizes an allegation, not a judicial conclusion. The plaintiffs still need to establish what information moved between models and whether copyright law reaches that transfer.

The technical evidence will matter enormously. A court could examine datasets, training logs, model evaluations, internal documentation, and testimony from Suno’s engineers.

It may also need to distinguish general musical capability from protectable expression. Copyright does not grant ownership over a genre, broad style, production method, or abstract musical idea.

Universal and Sony must therefore show more than similarity to popular music. Their case becomes stronger if they can trace identifiable protected material through specific technical processes.

Suno can answer that model behavior is not a stored catalog. It can also argue that learning statistical relationships from music serves a transformative purpose protected by fair use.

The company has previously described training on copyrighted music as lawful. Its 2024 response framed the labels’ litigation as an attempt to suppress new competition.

That defense remains central because copyright law does not contain a special rule for generative music. Courts must apply existing doctrines to development practices that do not resemble conventional copying in every respect.

Fair use will likely turn on several contested facts. These include why Suno copied recordings, how much it used, what its outputs contain, and whether the service harms licensing markets.

The labels’ recent commercial agreements strengthen one portion of their narrative. They argue that Suno’s own deals show a functioning market for authorized AI training.

Suno reportedly says its partnerships concern broader collaboration and new revenue opportunities, not simply payment for training data. That distinction could affect how a court evaluates market substitution.

The competing stories are clear. The labels describe a market that Suno bypassed before selectively entering it under legal pressure.

Suno describes a new creative technology whose learning process differs from distributing or replaying the source recordings. It says collaboration can develop without conceding that earlier training was unlawful.

The v6 architecture adds another layer. Even if Universal and Sony recordings never entered its direct dataset, the court could consider whether other inherited signals have legal significance.

Copyright doctrine offers no simple contamination test for model lineage. A later model is not automatically infringing because its predecessor faced allegations.

Conversely, deleting original files does not necessarily eliminate liability for completed copying. It also may not cure downstream products if those products reproduce protected material.

The judge will need evidence connecting those principles to Suno’s actual systems. Marketing descriptions such as “built from the ground up” cannot substitute for a technical record.

The dispute could influence how other AI developers document clean-room transitions. A clean-room process separates new development from disputed materials and personnel knowledge through defined controls.

Developers might record dataset provenance, prohibit old-model outputs from new training sets, and isolate preference data by model generation. They could also document which evaluations and tools cross that boundary.

Those safeguards carry costs. They can discard useful feedback, slow development, and reduce continuity for users who expect a new model to improve on its predecessor.

Yet weak documentation creates another risk. A company may claim a clean reset while lacking evidence that its data and model lineage support that claim.

Suno’s legal exposure therefore depends on more than whether v6 contains Universal or Sony files. It depends on how courts characterize the information that traveled from earlier systems into the new one.

The $9 Billion Figure Is a Ceiling, Not a Verdict

The largest number in the story measures maximum statutory exposure, while the case’s real importance lies in liability and injunctive relief.

The theoretical calculation is straightforward. The labels assert 60,202 works, and the maximum statutory award for willful infringement is $150,000 per work.

The legal path to that maximum is not straightforward. Universal and Sony must first prove ownership, copying, actionable infringement, and any facts supporting willfulness.

Suno can dispute the works list, challenge causation, invoke fair use, and argue against maximum damages. Courts also retain discretion when selecting awards within the statutory range.

The two cases may interact. The first action addresses 560 representative recordings and is moving toward a decision on Suno’s fair-use defense.

A ruling favorable to Suno could weaken the second case’s central copyright theory. A ruling favorable to the labels could strengthen their position across the expanded catalog.

Appeals could delay final resolution. Settlements or licensing agreements could also change the parties and remedies before any damages trial occurs.

The labels seek an injunction as well as money. An injunction is a court order requiring a party to stop specified conduct, and it can matter more than a headline damages estimate.

A broad order affecting models, retained datasets, or derived materials could force technical changes across Suno’s service. A narrower order might address particular recordings or development practices.

The complaint also asks for a jury trial and legal costs. None of those requested remedies should be treated as granted at this stage.

Suno’s financial position gives the dispute commercial urgency. The company announced a $400 million funding round in June 2026 at a $5.4 billion valuation.

That valuation is not cash available for a judgment. It reflects investor expectations about the company’s future rather than a legal assessment of its assets or liabilities.

According to a funding report, Suno had raised more than $819 million in total financing by September. The same report noted continuing investor confidence despite the litigation.

Large funding rounds can help a company absorb legal costs and build licensed products. They can also make it a more attractive target for plaintiffs seeking meaningful remedies.

The user base creates further stakes. Suno says more than 100 million people have used its service, while external reporting has described millions of daily song generations.

Company usage figures require caution because definitions can vary. A registered account, one-time visitor, active creator, and paying subscriber represent different levels of adoption.

Still, Suno has clearly moved beyond a laboratory experiment. Its models support real creative workflows, social content, demos, personalized songs, and material uploaded to streaming platforms.

That scale is central to the labels’ market-harm argument. They say machine-generated tracks compete for discovery, listener attention, and shares of fixed streaming royalty pools.

The complaint cites Deezer data indicating that AI-generated music accounted for more than half of daily track deliveries by July 2026. It places the daily volume near 90,000 tracks.

Volume alone does not prove infringement by Suno. AI tracks can originate from many tools, and some use authorized or original material.

However, the figure illustrates why rightsholders now view generative music as a distribution problem as well as a training dispute. Cheap production can increase catalog supply faster than platforms can review it.

Streaming services must distinguish legitimate creative use from impersonation, fraud, spam, and unauthorized derivatives. Labels want AI developers to bear more responsibility before those tracks reach distribution.

Suno says technical measures can help. It has discussed watermarking generated audio and limiting mass downloads, while arguing that distributors must govern content accepted onto their platforms.

Those measures address downstream abuse more directly than training rights. They cannot resolve whether Suno lawfully copied recordings during model development.

The skeptical view should therefore cut both ways. The $9 billion framing can exaggerate the likely financial outcome, but Suno’s licensed pivot does not settle the underlying claims.

Universal and Sony have not proved that v6 contains protected expression from their recordings. Suno has not publicly provided enough technical evidence to validate a complete separation from earlier systems.

A court record offers the right venue to test those positions. Until discovery develops, confident claims about either contamination or cleanliness remain premature.

What the Suno Case Will Test Next

Three developments will determine whether the lawsuit becomes a turning point for AI training or another dispute resolved through commercial licensing.

The first signal is the fair-use decision in the original 2024 case. That proceeding involves fewer works, but it contains the foundational argument about training on copyrighted recordings.

A victory for Suno would strengthen the view that model training can qualify as transformative use. It would also weaken the new case before its larger catalog receives extensive discovery.

A victory for Universal and Sony would increase pressure on Suno to settle or license additional catalogs. It would also encourage other rightsholders to pursue claims against AI developers.

The result will depend on the facts before the court, so it may not create a universal rule. Other models use different datasets, controls, output filters, and commercial structures.

The second signal is discovery concerning v6 model lineage. The parties will likely focus on synthetic outputs, preference records, distillation, retained copies, and internal descriptions of the rebuild.

Clear evidence that Suno isolated v6 from disputed expression would weaken the labels’ inheritance theory. Evidence showing extensive transfer from earlier models would strengthen it.

This inquiry matters far beyond music. Developers of image, video, text, and code models frequently reuse evaluations, synthetic data, and feedback from previous generations.

A ruling that treats those resources as carrying copyright liability could change model migration practices. Companies would need stronger provenance records and stricter boundaries between training generations.

The opposite result could support a practical route out of historical data disputes. Developers might replace models using licensed or controlled datasets while retaining general product knowledge.

The third signal is whether Universal and Sony negotiate. Warner’s agreement with Suno shows that litigation and partnership can become consecutive stages of the same commercial strategy.

Universal has already reached AI music arrangements elsewhere, while Sony continues exploring authorized uses of its catalog. Neither company rejects the technology in every form.

A Suno agreement would weaken the idea that this dispute must end with a definitive court judgment. It would instead reinforce licensing as the industry’s main route forward.

No agreement, combined with aggressive discovery, would suggest that the labels want precedent on fair use and model inheritance. That outcome would prolong uncertainty for developers and investors.

Users should also watch how Suno handles older models. Faster retirement would support its effort to move the service onto a licensed foundation.

However, replacing those models without satisfying creators can reduce retention and trust. The company must balance legal separation against product quality and workflow continuity.

For developers, the practical lesson is already visible. Dataset licenses alone may not protect a new model when teams reuse synthetic outputs, preference signals, or predecessor behavior.

For enterprise buyers, vendor diligence now needs to cover model lineage. Contracts should address training provenance, derived datasets, indemnity, audit rights, and the treatment of previous model generations.

For musicians, the case could shape whether opt-in systems become meaningful sources of control and compensation. It could also define how far labels can pursue claims involving technical inheritance.

For ordinary AI music users, immediate access is unlikely to disappear because a complaint was filed. Yet model availability, download policies, output controls, and commercial-use terms can change during litigation.

The Suno copyright lawsuit is therefore not best understood as a confirmed $9 billion judgment. It is a test of whether licensed AI can make a clean break from contested development.

Watch the original fair-use ruling first, v6 discovery second, and licensing negotiations third. Together, those signals will show whether Suno’s reset survives legal scrutiny or requires another redesign.