Elvis music rights holders and Eminem's publisher file two new lawsuits against Suno

Suno is one of the most prominent AI services that turns a text description into a complete song with vocals. Now the company, valued at $5.4 billion, faces two more lawsuits. Elvis Presley's rights holders and Eminem's publisher filed them on November 6 in federal court in Northern California. They join other rights holders already suing Suno.

What matters is not just the plaintiffs' names, but how their claims are structured. TNW reports that the lawsuits raise two distinct complaints: one about songs and another about voice. We do not have the full text of the lawsuits, so what follows is an analysis of the reasoning, rather than an account of the case filings.

Songs and voice: two separate legal claims in one

The first claim will be familiar to anyone following AI and copyright. Musical works are protected, and rights holders want to know which songs the model was trained on and whether it reproduces portions of them. This is a dispute about training data and what constitutes copying.

The second claim concerns voice. In the United States, a performer's voice is protected by more than copyright law. Rights of publicity protect likeness and voice, and Tennessee's ELVIS Act, which protects voice and likeness against AI imitation, has been in force since 2024. It appears in the URL of the original article, so the voice-related claim is likely built around it. That is our inference, not a quotation from the lawsuit.

Why this matters: catalog rights can be covered by a licensing agreement, as labels are doing with AI companies. But a song license does not answer whether a service may synthesize a voice resembling a particular artist's. These are two separate risks, and each needs to be addressed individually.

What this means in a simple example

Imagine you are building an app that generates jingles and background music. A user enters the prompt "sing like Elvis." Even if the model was trained on licensed data and does not reproduce any song verbatim, the question of voice remains. Liability could extend beyond the model developer to the product that allowed the request and presented the result to the user.

That has practical implications. Some teams will introduce prompt filters for artists' names; others will check generated audio for similarities to well-known voices. Others will need to rethink how they label generated content and what they put in their terms of service.

Limitations and practical takeaways

First, some caveats. A lawsuit is not a court ruling. The outcome of these cases is unknown, and disputes over AI music often end in settlements and licensing deals rather than court judgments. We have also not seen the complaints themselves and do not know which legal provisions and facts each plaintiff relies on.

Still, the signal to the industry is clear:

• Audio generation developers should document the provenance of their training data separately from their safeguards against voice imitation.
• Product managers should add restrictions on prompts requesting specific artists' styles to their backlogs while the rules remain unclear.
• Founders should treat legal risk as part of their unit economics, especially when a multibillion-dollar valuation attracts more plaintiffs.

The court in Northern California will not settle everything at once. But a case that separates voice and songs into distinct claims offers a glimpse of how future lawsuits against any AI service working with music will be structured.