Elvis Presley's Estate and Eminem's Publisher Sue Suno Over Style Cloning

If you build a product on a generative model, you probably have a list of banned words in prompts. The new lawsuits against Suno show that such a list alone may not be enough.

What happened

Suno, one of the most prominent music generation services, has received two lawsuits at once. In the first, Elvis Presley's rights holders claim the company violated Tennessee's ELVIS Act (Ensuring Likeness Voice and Image Security). The law makes it illegal to clone a performer's voice using AI. In the second lawsuit, the publisher of a number of Eminem's biggest recordings alleges that the service generates music and lyrics that imitate the artist.

Presley's lawyers write quite vividly: according to them, Suno "mechanically encoded" the identities, labor and time of artists and turned them into a conveyor belt for profit. The complaint goes on to say that Suno's "race to the bottom" seeks to make human creativity unnecessary, reproducing without a license in seconds what takes people years.

How these lawsuits differ from earlier ones

An important detail: the new lawsuits resemble the September lawsuit by Jason Isbell and others. That suit claimed Suno created imitations of the plaintiffs, but did not directly accuse the company of copyright infringement. In other words, lawyers are increasingly approaching the issue not through classic copyright, where you must prove the model was trained on specific tracks, but through the right to voice, likeness and name.

Why is this convenient for plaintiffs? The similarity of a voice or style can be demonstrated in the model's output. There is no need to dig into what was in the training data. It is enough to play two tracks for the jury and ask: do you hear it?

What Suno says

Company representative Kat McCormack told The Verge that Suno was created to help people make new original music, not to trade on someone else's name. According to her, the service does not allow artist names and song titles in prompts, and also works with third-party providers that check uploaded audio files and lyrics for possible abuse. The company promises to keep strengthening its safeguards.

The problem is that an input filter doesn't protect against everything. A user can describe a voice and manner in words without naming the artist, and the model will still produce something recognizable. The lawsuits target precisely the output, not the wording of the request.

What this means in practice

For developers and product managers working with audio, voice or image generation, the takeaway is simple.

First, check the output, not just the input. A prompt filter is necessary, but not sufficient. If possible, compare results against known voices and works.

Second, take regional laws into account. The ELVIS Act applies in Tennessee, but similar initiatives are appearing in other jurisdictions. If you have users in the US, the right of publicity over voice becomes a separate risk area.

Third, think through how you handle uploads. Features like "upload a reference" are convenient, but they are the most common route through which other people's material slips in.

Limitations

So far these are only lawsuits. The court has not yet decided whether Suno broke the law, and the company's position may prove persuasive. It is also unknown how courts will assess "style imitation": style by itself is usually not protected by copyright. But Sony, UMG, Isbell, Reservoir Media and others have already sued Suno, so the company will have to fight on several fronts at once.

The main takeaway for the industry: legal pressure on generative services is shifting from the question "what did you train on" to "what do you output." Safeguards need to be built in now, before the first letter from lawyers arrives.