Rights holders for Elvis Presley's recordings and the publisher of some of Eminem's biggest records filed two lawsuits against Suno on 8 November 2026, as The Verge reported. The claim: the company generates music and lyrics mimicking those artists without a licence. In the Presley suit, lawyers rely not on copyright but on a Tennessee state law protecting voice and likeness, known as the ELVIS Act.
Key takeaways
- Two new suits: from rights holders for Elvis Presley's recordings and from the publisher of some of Eminem's records
- The legal basis in the Presley case is the Ensuring Likeness Voice and Image Security (ELVIS) Act — a Tennessee law barring AI companies from cloning a singer's voice
- Suno says it does not allow prompts naming specific artists or songs and screens uploaded audio files and lyrics through third-party providers
- The new suits join earlier actions brought by Sony, Universal Music Group, Jason Isbell and Reservoir Media
- In September 2026 Suno released its v6 model family, which the company says was trained on licensed data from Warner Music Group, BMG and Believe
Picture the following situation
A user of a music generator types a prompt describing a track: tempo, mood, vocal type, a 1950s genre. The system returns a finished track — with vocals that sound recognisably like one specific, long-dead artist.
Nobody typed that artist's name. No recording of theirs was used as input. And yet the result is similar enough that a listener will associate it with one person.
But a problem appears: if the output contains no copied fragment of a recording, what exactly has been infringed? Copyright protects a work, not a timbre of voice.
And that is the gap the statute invoked in the Presley case is meant to close.
What are the new suits built on?
Both suits share the same construction: the claim concerns not the copying of recordings but the appropriation of artistic identity.
Lawyers representing the holders of rights to Presley's recordings argue that Suno violated the Tennessee statute titled Ensuring Likeness Voice and Image Security — the ELVIS Act for short. As The Verge describes it, the provision makes it illegal for an AI company to clone a singer's voice.
The language of the complaint is deliberately sharp. The lawyers write that “Suno has mechanically encoded itself with artists' identities, hard-work, creativity, and time, and harnessed them all for a consistent conveyor belt of cold hard cash.”
Later in the complaint comes a second charge — about scale and speed:
Suno's ravenous race to the bottom aims to make human creativity obsolete, coldly replicating without a license in seconds what takes years to learn, struggle with, maintain, and endlessly attempt to make better and perfect.
An excerpt from the suit filed by the holders of rights to Elvis Presley's recordings, quoted by The Verge.
Both suits, as The Verge notes, closely mirror the case brought by Jason Isbell and others in September 2026. That complaint also alleged the creation of imitations of the plaintiffs, but did not claim copyright infringement directly.
| Plaintiff | Basis of claim | When |
|---|---|---|
| Holders of rights to Elvis Presley's recordings | ELVIS Act — cloning a singer's voice | 8 November 2026 |
| Publisher of some of Eminem's records | imitating an artist without a licence | 8 November 2026 |
| Jason Isbell and other artists | creating imitations, without a copyright claim | September 2026 |
| Sony and Universal Music Group | copyright infringement in model training | ongoing |
| Reservoir Media | infringement of catalogue rights | ongoing |
What do these names mean?
The dispute turns on terms straddling law and engineering that do not explain themselves.
- ELVIS Act — a Tennessee state law, in full the Ensuring Likeness Voice and Image Security Act. It protects voice and likeness as personal attributes, independently of rights in specific recordings. The equivalent of a right of publicity extended to vocal timbre.
- Prompt filtering — blocking queries containing certain names and titles before the model begins generating. The equivalent of a blocklist on the input side, rather than control over what came out.
- Guardrails — a protective layer around the model deciding what the system must not do. It operates alongside the model, not inside its weights.
- Licensed data — training material the company holds an agreement for with the rights owner. For Suno v6 these are, per the company, datasets from Warner Music Group, BMG and Believe.
Together these four elements show why the case is technically hard: even a model trained solely on licensed data can learn stylistic traits nobody licensed separately — because style cannot be extracted from a recording the way a single track can.
What does Suno say?
The company does not address the legal claims, describing protective mechanisms instead.
Cat McCormack, a Suno spokesperson, told The Verge that “Suno exists to help people create new, original music, not to trade on anyone's name or likeness”, and that the company prohibits unauthorised use of an artist's work.
The specific measures she lists: not allowing prompts for specific artists' names and songs, and working with third-party providers to screen uploaded audio files and lyrics for potential misuse.
A separate argument concerns the model layer itself. As TechCrunch reported on 9 September 2026, Suno then released its v6 model family and stated it is not trained on the data used for previous versions. The company had earlier settled with Warner Music Group and struck a deal with BMG in August 2026. In August it also announced watermarking for generated tracks.
Why isn't ordinary copyright enough?
The question is natural: if the company trained a model on other people's recordings, a copyright suit should suffice. Such cases are in fact already running — brought by Sony and Universal Music Group.
The difficulty is that copyright answers a different question. It protects a specific work and a specific recording. If a generated track contains no copied fragment, infringement must be proved at the level of training rather than output — and that is a dispute over the legality of model training itself, which remains unresolved.
A statute like the ELVIS Act sidesteps this. It does not ask whether something was copied, only whether a specific person's voice was cloned. That shifts the burden of proof from comparative analysis of recordings to the recognisability of the effect. Which is why Isbell's September suit did not claim copyright infringement directly — and why the new suits repeat that construction.
There is a second problem neither route resolves: responsibility for the output. Prompt filtering blocks a name on the input side, but does not stop a description that reaches the same name by a roundabout path.
Is it working already?
No — these are filed suits, with no ruling. The two new actions were lodged on 8 November 2026 and join proceedings brought by Sony, Universal Music Group, Jason Isbell and Reservoir Media. None of them has a judgment yet.
What is confirmed: the existence of both suits and the content of the quoted excerpts, reported by The Verge, plus the official response from Suno's spokesperson. Also confirmed is that since September 2026 Suno offers a v6 model family the company describes as trained on licensed data.
What is not known: whether the ELVIS Act will cover Suno's activity. This is a first line of argument, not settled case law, and the reach of a state statute against a globally operating company is a separate legal question.
Why does this matter?
The centre of gravity is moving from “what was the model trained on” to “what can the model reproduce”. For builders of generative systems that is an awkward shift, because training data can be documented while resemblance of an output to one person's style cannot. If this line of argument holds, licensing datasets will stop being a sufficient defence.
What next?
- Both suits await hearing — the decisive point will be whether a court finds the ELVIS Act applicable to generating vocals without using any recording of the specific person
- Suno has announced an opt-in programme with labels allowing artists' songs to be used for generative features — its terms will decide whether some disputes convert into licences
- Identified risk: filtering prompts by name does not block descriptions leading to the same effect, so the declared safeguards do not answer the suits' central charge





