Suno AI Lawsuits: Elvis, Eminem and Artist Rights
Elvis and Eminem rights holders are suing Suno over AI music clones. Here’s what the claims mean for artist rights, safeguards and AI music businesses.
The Suno Lawsuits Put AI Music's Guardrails to the Test
Suno is facing two more lawsuits over AI-generated music. One suit, brought by rights holders tied to Elvis Presley's work, and another by a publisher behind some of Eminem's best-known records, allege that Suno makes tracks and lyrics that imitate those artists. In the Presley case, the plaintiffs' lawyers say the alleged conduct violates Tennessee's ELVIS Act, a law aimed at unauthorized AI cloning of a person's voice and likeness.
These complaints add to a growing pile of legal trouble for Suno, including lawsuits involving major record companies, Jason Isbell, and Reservoir Media. And while the outcome in court will matter, these cases also test a tougher question: when a system can produce a convincing imitation, how do existing rights actually apply - and do companies' safeguards hold up in practice, not just in policy?
Suno denies any intent to profit off artists' identities. The company says its goal is to help users create original music, that it blocks prompts that name specific artists and songs, and that it screens uploaded audio and lyrics for potential misuse. Still, those claims and the plaintiffs' allegations are in dispute. No court has yet found that Suno violated the law in these cases.
This article breaks down what the lawsuits claim, how voice-and-likeness claims differ from copyright claims, what Suno says it does to reduce misuse, and what the disputes could mean for AI music businesses - and the people whose work inspires them.
What do the Elvis and Eminem lawsuits allege?
Both complaints center on alleged AI-generated imitations of well-known artists. In the Presley case, the plaintiffs' lawyers argue that Suno uses artists' identities and creative work - along with the effort behind it - to generate music at scale without permission. Their filing describes Suno's output as a "consistent conveyor belt" of commercial releases. They also claim the system can reproduce, in seconds, qualities that artists say they took years to develop.
The Presley lawsuit also has a specific Tennessee-law angle. The lawyers for the rights holders say Suno violated the Ensuring Likeness Voice and Image Security Act, commonly called the ELVIS Act. The law covers unauthorized use of a person's voice or likeness, including through AI. Whether the alleged facts fit the statute is something the legal process will have to decide.
The second suit comes from the publisher of some of Eminem's major records. According to reporting, that publisher likewise alleges that Suno generates music and lyrics that mimic the rapper. But the public reporting so far doesn't spell out the full scope of the claims or the specific remedies being sought, so it's not safe to treat these lawsuits as identical in legal detail.
The complaints also resemble an earlier lawsuit filed by Jason Isbell and other plaintiffs. That case, according to reporting, claimed Suno created imitations but did not directly focus on copyright infringement. The point matters: arguing that a system copied something in a recording is not always the same thing as arguing that it generated sound tied to a performer's identity.
Allegations are not findings
A lawsuit lays out one side's story and legal theory. It doesn't prove the system produced the challenged outputs, it doesn't automatically establish legal responsibility, and it doesn't settle what the law means in this context. Suno spokesperson Cat McCormack told The Verge that the company doesn't aim to trade on anyone's name or likeness and prohibits unauthorized uses of an artist's work.
For the cases, the practical questions will likely turn on what users can actually prompt the system to generate, what controls Suno had in place, how those controls work in real conditions, and what evidence ties the alleged outputs to protected rights. None of that can be settled by the language of a complaint or a company statement.
Key takeaways: These are allegations, not verdicts. The Presley complaint points to Tennessee's ELVIS Act, while the wider dispute raises the question of whether AI-generated imitation can violate rights even when a lawsuit isn't framed as copying a specific recording.
How are voice and likeness rights different from copyright?
Copyright usually protects specific creative works - like a recording, a composition, or a set of lyrics. A voice-and-likeness claim raises a different issue: whether someone used a person's identity without authorization. The two can overlap, but they don't work the same way.
For example, a copyright dispute might ask whether a model was trained on protected recordings or whether an output reproduces a recognizable part of a song. A voice-and-identity dispute might instead focus on whether generated audio imitates a performer's voice or persona in a way the law forbids. Which test applies depends on the relevant law and the facts. Even a voice that sounds "right" isn't automatically a legal problem by itself.
That distinction matters for generative AI because the output may be newly assembled rather than a direct copy of any one track. A user can ask for music that matches broad characteristics - say, a time period, mood, or instrumentation - without naming a performer. But if the result allegedly evokes a specific artist, the rights question may not come down to whether it contains a copied sample or lyric.
The ELVIS Act gives the Presley complaint a state-law framework to argue from. It's aimed at unauthorized AI-generated use of a person's voice and image. But the claim that Suno violated it is still an argument, not a settled interpretation of how the law applies to this product and these outputs.
Why the distinction matters to product teams
For an AI music company, focusing only on copyright can leave you with a gap. A system might block known song lyrics or audio files while still generating something that looks like an imitation when a user uses a descriptive prompt. On the other hand, a "this sounds like that" complaint doesn't, by itself, prove a specific output is unlawful. Good risk analysis has to separate inputs, model behavior, outputs, and how users monetize or distribute what they generate - rather than treating "AI music" as one single legal category.
A practical review asks:
- What is being protected? A recording, composition, lyric, voice, image, or broader identity can raise different issues.
- What action is alleged? Training, prompt handling, output generation, distribution, and monetization are different parts of the product lifecycle.
- Who made the request? User intent and company controls can both matter, depending on the claim.
- What evidence exists? If the argument is resemblance, it has to be tested against the actual prompt, the actual output, and the relevant rights.
This won't automatically answer legal questions. But it helps teams avoid the common mistake of assuming that one license, one filter, or one policy covers everything that could come up.
Are Suno's safeguards enough to prevent artist imitation?
Suno says it doesn't allow prompts that include specific artists' names and songs. It also says it uses third-party providers to screen uploaded audio and lyrics for potential misuse, and that it keeps strengthening its safeguards. Those are real-sounding controls - but a policy statement alone doesn't show how well they work across the messy reality of how people actually use the product.
Blocking named artists can catch straightforward requests. What's less clear, based on the information available, is how the system handles indirect prompts: misspellings, combinations of distinctive traits, or repeated attempts to get around a refusal. Screening uploaded audio and lyrics tackles one route to misuse, but it doesn't necessarily address what happens when someone describes a target sound without providing source material.
None of this proves the safeguards fail. It does suggest the evaluation should focus on behavior, not just the presence of rules. In a production service, controls need monitoring, testing, and a clear way to respond when users find a gap. The real question is whether protections still work under ordinary use - and under adversarial use - and whether the company can keep updating them as new patterns show up.
A useful safeguard test
AI music teams can assess music-generation controls across four stages:
- Before generation: Does the product detect prohibited artist names, song titles, and likely variations? Can users report a concern or request review?
- During generation: Does the system refuse, redirect, or limit requests that look like they target a specific artist? Do repeated attempts get handled differently from the first try?
- After generation: Can outputs be checked for potentially problematic similarities? Is there a process to investigate reports and restrict access to disputed material while a review is underway?
- Across the product: Are decisions logged well enough to understand what happened, and are controls tested after model or product changes?
These are operational questions, not a claim about Suno's internal systems. Public reporting describes the safeguards Suno says it uses, but it doesn't provide performance data or a full look at how reviews actually work.
There's also a trade-off. Broad filters can block legitimate prompts that reference genre, influence, or historical style. Narrow filters can miss attempts to reproduce an individual artist indirectly. A workable system needs a clear boundary, a review path for ambiguous cases, and regular testing against real use. The goal isn't to promise that imitation will never happen - it's to reduce foreseeable misuse and respond when controls fall short.
What could the cases mean for AI music businesses?
For AI music companies, the immediate risk isn't limited to whatever happens in court. Multiple lawsuits can raise legal costs, complicate licensing talks, and make customers, artists, and business partners more cautious. The cases also pressure a core product decision: whether a service should let users ask for something that resembles a specific artist, even if that demand drives usage.
A platform can still be commercially useful by supporting broad creative exploration without offering direct artist cloning. That choice shapes product design. Companies can offer controls for genre, arrangement, tempo, and mood - while tightening limits on requests intended to reproduce someone's voice or identity. What counts as "enough" legally will depend on the facts and the law, but the simplest path is to make the boundary explicit instead of relying on an unclear prohibition and hoping for the best after a problem appears.
Rights holders face their own practical challenges. They have to identify the outputs, document the relevant rights, and decide whether to pursue a claim, seek a license, or negotiate safeguards. A lawsuit can set a position, but litigation can be slow and expensive. In many cases, licensing or technical agreements may be more useful than fighting over every individual output. That's a business choice - not a guaranteed endgame.
The scaling problem
Generative music changes the economics of imitation. Human-made imitation takes time and skill, while a model can generate many variations quickly. That can increase the volume of material rights holders have to monitor, and it can also multiply the number of borderline cases companies may need to review. The issue isn't just that generation is fast. It's that the cost of producing and distributing possible imitations can drop faster than the cost of identifying and resolving disputes.
So an operating model needs more than a terms-of-service clause. It needs a documented approach to permissions, prompt restrictions, output review, user reports, and escalation. It also needs a decision-maker who can change the product when evidence shows a control isn't working. For smaller companies, that may mean limiting features or launch markets until the review process is ready. Those limits have a cost - but so does shipping a capability that can't be governed at scale.
For artists and publishers, these cases are a reminder to be clear about what they want protected and what outcome they're aiming for. For AI companies, they're a warning against treating "the user made the prompt" as the whole risk strategy. The product still determines what users can ask for, what gets generated, and how complaints are handled.
Conclusion
The lawsuits against Suno are part of a broader test of how rights apply to generative music. The Presley case brings Tennessee's ELVIS Act into focus, while both complaints raise a practical question for the industry: how should a product respond when users try to get music that evokes a specific artist?
The answer depends on the evidence, the applicable law, and what the courts decide. In the meantime, AI music businesses should treat safeguards like part of their operations - not a checkbox. That means defining what's restricted, testing how users can push at the limits, documenting decisions, and having a process for responding when a complaint seems credible. This won't eliminate uncertainty, but it's likely more durable than assuming a prompt filter alone settles the issue.