Lawsuits against AI music generators are moving beyond copyright. Claims built on voice and likeness rights may be harder for these companies to defend, because they target what the software outputs rather than how it was trained.
Key takeaways
- The Verge reports that rights holders for Elvis Presley’s music and a publisher of some of Eminem’s best-known records have filed two new lawsuits against the AI music company Suno.
- According to The Verge, the Presley lawsuit relies partly on Tennessee law, not just federal copyright, to challenge music that imitates the artist.
- Copyright generally protects specific recordings and compositions rather than a musical style, so it can struggle with AI tracks that sound like an artist without copying a particular song.
- Voice and likeness laws cover an artist’s identifiable traits, which gives rights holders a way to challenge imitation even when no specific work has been copied.
The argument: imitation, not copying, is the weak point for AI music
Most early legal pressure on generative AI music asked whether companies could lawfully train models on copyrighted recordings. That question matters, but it is hard to settle. It depends on how training works technically, on whether training counts as a transformative use, and on evidence about datasets that the companies hold and the plaintiffs often cannot see.
The two new lawsuits against Suno point to another route. The Verge reports that the rights holders say Suno generates music and lyrics that mimic particular artists. In the Presley case, the lawyers also claim a breach of Tennessee law. That moves the dispute away from the inputs a model learned from and towards the outputs it produces.
The argument here is that this shift is significant. A system that can be prompted to produce something recognisably in the voice or manner of a specific performer is exposed in a way that a dispute over training data does not capture. Voice and likeness protections were written for exactly this kind of harm: someone’s identity used without permission. Generative audio tools make that harm cheap and repeatable. If courts accept that framing, AI music companies may have to limit what their systems will produce, not only what data they collect.
This is an analytical position, not a prediction about these particular cases. The Verge’s report does not explain how either lawsuit is structured beyond the outline above. Their outcome is unknown, and the details of the claims, the remedies sought and Suno’s response are not set out in the material available here.
Evidence one: copyright has a gap where style lives
Copyright law in most jurisdictions protects particular expressions, such as a recording, a melody or a set of lyrics. It does not usually protect a genre, a vocal manner or a general sound. A new song that captures the feel of a mid-century rock and roll singer, or the delivery of a well-known rapper, may infringe nothing if it reproduces no protected melody, lyric or recording.
Generative music tools work in exactly that space. Their main appeal is producing new material on demand in a requested style. A rights holder who hears an AI track that sounds unmistakably like their artist may find that copyright gives them little to act on unless they can identify specific copied elements.
That gap explains why plaintiffs would look for other legal grounds. The Verge’s report that the Presley suit invokes Tennessee law, rather than relying only on copyright, fits that pattern. Whatever the merits of the individual claims, choosing to build a case on state law suggests the plaintiffs see copyright alone as insufficient for the harm they describe.
The Eminem-related suit, brought by a publisher, according to The Verge, is described as covering both music and lyrics. Lyrics sit closer to traditional copyright, since text can be compared directly. The pairing shows how the two approaches can be used together: copyright where there is something concrete to compare, and identity-based claims where the complaint is about resemblance.
Evidence two: output claims are easier to demonstrate
Training-data claims depend on information the defendant controls. To show that a model was trained on particular recordings, plaintiffs usually have to infer it from the model’s behaviour or obtain internal records through litigation. Even then, the legal significance of training is disputed.
Output claims work differently. If a system can be prompted to produce a track that a listener would identify with a particular performer, that track becomes evidence that anyone can play. A judge or jury does not have to understand model architecture to hear whether a voice sounds like someone. The question becomes one of perception and identification, which courts already deal with in other areas.
This is why the mimicry framing in the lawsuits reported by The Verge matters. The complaint, as described, concerns what Suno generates. That puts the focus on what the product actually does in use. It also means a defence based on how training works may not answer the claim, because the claim is not about training.
None of this means output claims are easy to win. Plaintiffs still need to show that the resemblance is close enough, and specific enough to their artist, to count as using that artist’s identity. But the evidence is easier to understand and harder to dismiss as speculation about hidden processes.
Evidence three: state law has adapted to synthetic voices
Tennessee has a law that explicitly protects a person’s voice, alongside their name and likeness, against unauthorised commercial use. It was widely discussed as a response to AI-generated vocal imitations, and its subject matter suits a dispute involving a performer so closely tied to that state’s musical history. The Verge reports that the Presley suit claims Suno violated Tennessee law. The full name and the specific provisions relied on are cut off in the source material, so which statute is involved, and how it applies, cannot be confirmed here.
The broader point is that legislators have been adapting older publicity and likeness rights to synthetic media. Traditional right-of-publicity claims often centred on images and endorsements. Explicitly including voice, and treating the tools that produce voice clones as a possible source of liability, extends those rights into territory that copyright does not cover.
The Verge also describes these as “yet another pair” of suits against Suno, which indicates that legal pressure on the company is building. A growing number of cases, brought by different rights holders on different legal theories, raises the chance that at least one theory produces a ruling with wider consequences.
The case against: likeness law is patchy, contested and limited
The strongest objection is that voice and likeness rights are a weaker basis than they look. In the United States they mostly come from state law, and the states differ a great deal. Some protect voice explicitly and some do not. Some recognise rights after death and some do not. Protection after death matters for estates, and its length and scope vary between jurisdictions. A claim that works in Tennessee may have no equivalent elsewhere, and questions of which state’s law applies to a service used across the country could be complicated to resolve.
There are also free expression limits. Imitation, parody and homage have a long history in music. Sound-alike performances, tribute acts and genre revivals are ordinary parts of the industry. Courts tend to be cautious about letting identity rights restrict expressive work, and a defendant could argue that producing music in a recognisable style is protected expression, not commercial appropriation of a person.
A practical problem follows. Proving that a generated track uses a specific person’s identity, rather than a general style many artists share, may be difficult. Many performers share vocal traits. A defendant could argue that any resemblance comes from the user’s prompt, not from the product, or that the product only reflects widely shared conventions.
Finally, copyright still offers advantages that likeness law may not, including federal jurisdiction and well-established remedies. A large ruling on training data could change the economics of the industry more decisively than a series of state-level identity claims. On this view, copyright remains the main battleground and likeness claims are a supplement.
These objections are serious. They suggest that identity-based claims may succeed in some places and fail in others, producing an uneven picture rather than a clear limit on AI music.
What would change this conclusion
Several developments would weaken or strengthen the argument.
If courts dismiss the Tennessee-based claim early, for example on free expression grounds or because the law does not reach a tool provider as opposed to an individual user, the argument that likeness law is a sharper threat would lose much of its force.
If, on the other hand, a court lets such a claim go forward, or treats the ability to generate a recognisable voice as enough to raise liability, that would support the argument and probably encourage similar suits in other states with comparable laws.
A decisive ruling on training data under copyright would also matter. If courts found that training on recordings without a licence is itself unlawful, copyright would again become the dominant pressure on the industry, whatever happens with likeness claims.
Federal legislation creating a national voice and likeness right would remove the patchwork objection and make identity-based claims much more powerful. Without it, the uneven state-by-state picture continues.
Lastly, changes to the products themselves would be informative. If AI music companies began blocking prompts that name specific artists, or filtering outputs that closely resemble known voices, that would suggest they see output-based liability as a real risk. Whether Suno has made or will make such changes is not known from the material available.
Sources and further reading
- The Verge: reporting on the two lawsuits brought against Suno by music rights holders, including the reference to Tennessee law.
- Tennessee state legislature: published text of the state’s laws protecting name, image, likeness and voice.
- United States Copyright Office: guidance and reports on copyright and artificial intelligence, including the treatment of style.
- Academic legal journals: commentary on the right of publicity, rights after death and their interaction with free expression.
Surfaced from the rss:verge signal “AI music lawsuits”. AI-assisted draft, editorially reviewed.
