Lehrman v. Lovo: Right of Publicity

A federal court let voice actors' right-of-publicity and contract claims against an AI voice-cloning company proceed while dismissing their copyright theories. The decision maps the legal terrain for performers facing synthetic replicas of their voices.

Lehrman v. Lovo, Inc., No. 1:24-cv-03770 (JPO) (S.D.N.Y. July 10, 2025), is among the most closely watched cases at the intersection of generative artificial intelligence and performers’ rights. Voice actors Paul Lehrman and Linnea Sage allege that Lovo created and sold unauthorized AI “clones” of their voices, having obtained samples through the freelance marketplace Fiverr under pretenses that concealed the intended use. Ruling on Lovo’s motion to dismiss, Judge J. Paul Oetken granted the motion in part and denied it in part. The pattern of what survived and what fell is the decision’s lasting contribution. For Los Angeles’s community of voice actors, musicians, and performers, it is a foundational text.

At a glance

Case: Lehrman v. Lovo, Inc., No. 1:24-cv-03770 (JPO) (S.D.N.Y.)
Decided: July 10, 2025 (opinion and order on motion to dismiss), Judge J. Paul Oetken
Holding: Right-of-publicity, breach-of-contract, and consumer-protection claims survive, as does Sage’s copyright claim based on Lovo’s reuse of her actual recording; Lanham Act and the imitation-based copyright claims dismissed (the training claim with leave to amend)
Status: Stayed. Plaintiffs filed a Second Amended Complaint on July 31, 2025, and Lovo moved to dismiss the repleaded copyright training claim. Lovo then filed for Chapter 7 bankruptcy in May 2026, and on June 11, 2026 Judge Oetken stayed the case and adjourned all deadlines, including a June 16, 2026 argument on that motion. As of July 2026 the stay remains in place and the motion is undecided.

What survived, and what did not

The court’s disposition divides cleanly along a doctrinal fault line.

Survived (motion denied): the plaintiffs’ claims under New York Civil Rights Law §§ 50–51 (the state’s right-of-publicity and voice-misappropriation regime); their breach-of-contract claim; their New York consumer-protection claims under General Business Law §§ 349–350; and, in the one copyright claim to survive, Sage’s claim that Lovo reproduced part of one of her actual copyrighted recordings in 2020 presentations and in videos on Lovo’s YouTube channel.

Dismissed: the Lanham Act false-association and false-advertising (trademark) claims; and most of the federal copyright claims, including the theory that training the AI on the plaintiffs’ existing recordings infringed (dismissed with leave to amend) and the theory that the cloned outputs were infringing reproductions or derivative works. The court also dismissed the common-law claims other than breach of contract (fraud, conversion, unjust enrichment, and unfair competition), holding the conversion and unjust-enrichment theories preempted by the Civil Rights Law.

That split is the holding’s intellectual core: copyright recedes, and the state-law right of publicity advances.

Why copyright is the wrong tool for a voice

The dismissal of the copyright theories rests on a fundamental limit of the statute. Copyright protects original works of authorship fixed in a tangible medium, including the plaintiffs’ actual sound recordings. It does not protect the abstract qualities of a voice: timbre, cadence, accent, the ineffable identity that makes a voice recognizable. A new recording that imitates or simulates a person’s voice, without copying the fixed sounds of an existing recording, does not infringe the recording copyright, because what it appropriates (vocal identity) is not the subject matter of copyright at all.

The statute says so expressly: the sound-recording owner’s duplication and derivation rights “do not extend to the making or duplication of another sound recording that consists entirely of an independent fixation of other sounds, even though such sounds imitate or simulate those in the copyrighted sound recording.” 17 U.S.C. § 114(b).

That is why the “cloned output” theory failed, and why the training-infringement claim was, at least as pleaded, dismissed (the court faulted the plaintiffs for not explaining how the training process itself infringed, and gave them leave to replead). The one copyright claim that survived proves the rule rather than undercutting it: it rested on Lovo allegedly reusing Sage’s actual recording, an ordinary act of reproduction that had nothing to do with cloning. Otherwise the harm the plaintiffs describe is not the copying of their recordings but the appropriation of their identities. Identity is the province of the right of publicity, not copyright.

Why the right of publicity is the right one

New York’s §§ 50–51 prohibit the use of a living person’s name, portrait, picture, or voice for advertising or trade purposes without consent. The court held that this regime reaches AI-generated voice clones used commercially, rejecting Lovo’s argument that the statute’s digital-replica provisions excluded the plaintiffs. Construing the statute to encompass technological developments, the court allowed the voice-misappropriation claim to proceed. The court also rejected Lovo’s Statute of Frauds defense at the pleading stage, reasoning that because Lovo’s Fiverr messages were sent from accounts identified with people now known to be its agents, Lovo is bound by them even though its representatives did not sign their real names, and that Fiverr’s Terms of Service could supply any missing material terms as a connected writing. Whether those writings ultimately satisfy the Statute of Frauds, the court said, is a fact question not resolvable on a motion to dismiss, so the breach-of-contract claim stays alive.

The doctrinal lesson generalizes beyond New York. Performers confronting unauthorized synthetic replicas of their voices will, in most cases, find their remedy in state right-of-publicity law and in contract, not in the Copyright Act. That allocation has practical consequences, because right-of-publicity law is state law: it varies in scope, in its treatment of post-mortem rights, and in its interaction with the First Amendment, producing a patchwork rather than a single federal standard.

Open questions

Several questions remain live. First, whether the plaintiffs can successfully replead a copyright training claim (the dismissal was with leave to amend), and how such a claim would fare against the emerging fair-use framework of Bartz and Kadrey. Second, how right-of-publicity statutes that predate generative AI will be construed across jurisdictions, and whether courts will, as here, read them to encompass synthetic voices, or insist on legislative updates. Third, the looming First Amendment dimension: right-of-publicity claims must eventually be reconciled with expressive uses, an issue not resolved at the pleading stage. The wave of new state “digital replica” statutes will interact with these common-law and § 50–51-style claims in ways yet to be tested.

Implications for performers and AI companies

Performers: look to publicity and contract. The most durable claims against unauthorized voice cloning are state right-of-publicity and breach-of-contract theories, not copyright. Document the terms under which voice samples are provided.
AI companies: consent is the control. Building a voice product on samples obtained without informed consent for the actual use is the exposure the case identifies. Contractual permissions should match the intended use.
Counsel: mind the patchwork. Because publicity rights are state-specific, the analysis turns on which state’s law applies and on the scope of any applicable digital-replica statute.

Frequently asked questions

Can you copyright your voice? No. Copyright protects fixed sound recordings, not the abstract qualities of a voice. A recording that merely imitates a voice does not infringe the recording copyright, because 17 U.S.C. § 114(b) excludes independent fixations that imitate or simulate a protected recording. That is why the court dismissed the imitation-based copyright theories, while letting one claim proceed over Lovo’s alleged reuse of an actual recording.

What law protects against AI voice cloning, then? Primarily the state-law right of publicity (in New York, Civil Rights Law §§ 50–51, which covers voice), along with contract and consumer-protection law. Those claims were allowed to proceed.

Is the case over? It is not resolved on the merits, but it is currently halted. The July 2025 ruling was on a motion to dismiss, and the plaintiffs filed a Second Amended Complaint on July 31, 2025. After Lovo filed for Chapter 7 bankruptcy in May 2026, the court stayed the case on June 11, 2026. As of July 2026 the stay remains in place.