Does AI Voice Cloning Violate Publicity Rights?

If your product team told you your model can replicate a recognizable voice from a 30-second sample, do you treat that as a feature or as a liability?

Artificial intelligence has moved synthetic voice from demo to deployment. AI can now replicate human pitch and inflection with striking fidelity. What once required a recording studio, a sound engineer, and the performer’s cooperation can now be achieved with a trained model and a short audio sample.

That makes the question more than theoretical, as courts confront a pressing question: when does AI voice cloning cross the line into misappropriation?

The answer is becoming increasingly blunt. If you use an AI-generated voice in commerce without consent, especially one that evokes a specific person, you should expect to face publicity, unfair competition, or false endorsement claims. Copyright will rarely save you. Publicity law and trademark theories will do the real work.

Identity, Not Expression, Is the Real Battleground

The right of publicity, rather than copyright or patent, lies at the center of these disputes. It’s all about identity.

In the United States, publicity rights are state-based and uneven. Some states recognize a distinct statutory right; others fold the protection into privacy torts. But jurisdictions with mature entertainment economies like California, New York, Nevada, and Tennessee have made their position clear. Voices and digital replicas are protectable attributes of identity.

Recent corrective legislative updates include California’s AB 1836 and AB 2602, effective January 2025, which directly target unauthorized digital replicas of performers, living and dead. Illinois amended its Right of Publicity Act to clarify that digitally generated names, voices, and likenesses are protected property interests. New York expanded its Civil Rights Law in 2024 to address AI-driven deceptive practices.

If you commercially deploy a digital replica without consent, you are stepping into statutory damages and injunctive exposure. The compliance burden sits with the deployer, not the model.

Supported by Precedent

The foundational case remains Midler v. Ford Motor Co. When Bette Midler declined to sing for a car commercial, the advertiser hired a sound-alike and instructed her to imitate Midler’s voice. The public believed the Divine Ms. M was behind the mic.

The Ninth Circuit cut through the doctrinal noise. A voice is not copyrightable. But a distinctive voice is a protectable attribute of identity. If you imitate it because it has commercial value, that value belongs to the person.

That distinction, expression versus identity, has only sharpened in the AI context. Copyright protects fixed works. Publicity law protects the human signal embedded in them. Voice cloning does not collapse that boundary. It makes it harder to ignore.

AI Litigation Exposes the Fault Lines

A recent federal case, Lehrman v. Lovo, Inc., illustrates where plaintiffs are finding traction.

Two voice actors alleged that a defendant incorporated their voices into commercial AI-generated content after approaching them for limited internal use. The court found insufficient evidence that consumers believed the actors endorsed the platform to proceed on Lanham Act grounds. The voices functioned as components of the product, not as overt marketing signals.

The copyright claims failed for predictable reasons. No protected recording was shown to have been reproduced. Copyright protects fixed works, not the timbre of a throat.

But the complaint survived under New York’s Civil Rights Law. That statute prohibits the use of a person’s voice for advertising or trade purposes without written consent. The court also rejected a statute-of-limitations defense, treating each republication as a fresh violation.

That should get the attention of any platform running iterative releases. Ongoing deployment can mean ongoing liability.

Trademark Law Is Catching Up

Federal trademark law is also expanding into the AI perimeter. The Lanham Act prohibits uses likely to cause confusion about sponsorship or affiliation. Courts are increasingly willing to treat AI outputs as “use in commerce” when tied to commercial platforms.

In Advance Local Media LLC v. Cohere Inc., publishers alleged that an AI system generated hallucinated articles attributed to them, using their marks and mimicking their editorial structure. The court declined to dismiss. AI output, when deployed commercially and plausibly confusing, can trigger Lanham Act exposure.