The Problem: Your Voice Got Cloned and Federal Law Didn't Save You

You hear your own voice coming out of a synthetic mouth, saying things you never said, and your first instinct is to call a lawyer. That instinct is right, but the obvious legal theory probably won't work. A federal court in New York made that clear in Lehrman v. Lovo, Inc., and the ruling reshapes how voice cloning claims actually get brought.

The court dismissed the copyright claims outright. Under 17 U.S.C. § 114(b), a voice clone that mimics the sound of a recording is not a "copy" under the Copyright Act unless it duplicates actual recorded material. Trademark claims failed too, because the actors' voices were "tools of their trade" and lacked the secondary meaning needed to function as source identifiers. The judge warned that stretching the Lanham Act to cover any recognizable voice would turn it into a general right of persona.

The facts in Lehrman show why this stings. Voice actors were hired anonymously through Fiverr under assurances their recordings were for "academic research purposes only" and would not be disclosed externally. Their voices then surfaced commercially under pseudonyms like "Kyle Snow." One plaintiff had been paid $1,200, another $800.

So what survives? In Lehrman, the court let right of publicity claims proceed under New York's digital replica provision, along with breach of contract and consumer protection claims. The lesson is that federal IP law is a weak fit, while state law is where the leverage lives. Protection varies by state, and performer advocacy guidance directs people to check whether their own state recognizes voice in its right of publicity statute.

A second route is biometric privacy law. Journalists, podcasters, and audiobook narrators filed proposed class actions in Chicago federal court against nine companies, including Apple, Amazon, Meta, Microsoft, Nvidia, Samsung, Alphabet, ElevenLabs, and Adobe, alleging BIPA violations for using recorded voices as training data. Statutory penalties run $1,000 per negligent violation and $5,000 per intentional one. Prior BIPA settlements from the same plaintiffs' firm reached $75 million with BNSF in 2024 and $47.5 million with Motorola Solutions in 2025.

If you run a business deploying voice AI, the exposure cuts both ways. Vetting how your provider sources its voices matters as much as the technology itself. Answrr builds AI receptionist systems for service businesses, and compliance-conscious buyers can book a free setup call to see how consent-based voice deployment works in practice.

• Check whether your state's right of publicity statute covers voice, including digital replicas.
• If your recordings were scraped for training, biometric privacy statutes may offer statutory damages.
• Preserve every communication and licensing term from the original recording session.
• Ask vendors for licensing documentation, SOC 2 or ISO 27001 certification, and no-training-on-customer-data policies.

The Routes That Work: State Publicity Laws and Biometric Privacy Claims

If you want to sue because AI cloned your voice, federal copyright and trademark law will likely disappoint you. Two state-law routes, however, are proving viable: right of publicity statutes and biometric privacy claims.

The first route got a major boost in Lehrman v. Lovo, Inc., decided by a federal court in New York in July 2025. The court tossed the voice actors' copyright and trademark claims, but it let their right of publicity claims proceed under New York Civil Rights Law §§ 50–51, holding that the state's recently added "digital replica" provision could plausibly apply to AI voice clones even without a visual likeness. As the Fredrikson & Byron analysis put it, federal IP protections may not fit AI voice clones, but state publicity and consumer protection laws "could pose significant risks."

Protection isn't uniform, though. The National Association of Voice Actors advises performers to check whether their state has a publicity law covering voice, pointing to California's AB2602 and New York's S7676B as key examples.

The second route runs through Illinois' Biometric Information Privacy Act (BIPA). Journalists, podcasters, and audiobook narrators have filed proposed class actions in Chicago federal court against nine tech companies, including Apple, Amazon, Meta, Microsoft, Alphabet, and ElevenLabs, alleging the firms used thousands of hours of recorded voices to train AI without permission. Notably, these suits don't allege copyright violations at all; they rest on biometric privacy, publicity, and deceptive-practices theories, according to Reuters' coverage of the litigation.

The stakes are substantial:

• BIPA carries statutory damages of $1,000 per negligent violation and $5,000 per intentional or reckless violation, applied per victim and per subsection of the law.
• Plaintiffs' counsel asserts each of the nine defendants faces hundreds of millions of dollars in potential exposure, though that's their estimate, not a court finding.
• The same plaintiffs' firm, Loevy & Loevy, has already secured BIPA settlements of $75 million from BNSF in 2024 and $47.5 million from Motorola Solutions in 2025.

Defendants are pushing back. Apple's lawyers argue the plaintiffs "allege only that their recordings are available on the internet" without facts showing Apple actually used them for training. So outcomes remain unsettled.

For businesses deploying voice AI, these cases carry a practical lesson: sourcing matters. Whether you're hiring voice talent or choosing an AI receptionist for your service business, vet how the provider obtains and licenses voice data, and whether it holds security certifications like SOC 2 or ISO 27001. If you'd rather have that compliance review handled for you, you can book a free setup call with Answrr to see how a consent-based voice AI setup works in practice.

How to Build Your Case: Contracts, Riders, and Documenting Deceptive Procurement

If your voice ends up in an AI model, your paper trail matters as much as the legal theory. The Lehrman v. Lovo ruling showed that federal copyright and trademark claims over voice clones mostly fail, while contract and state publicity claims survived. That means the strength of your case often comes down to what you documented before and after the recording session.

Start by checking whether your state protects voice under its Right of Publicity statute. Protections vary widely, and the National Association of Voice Actors (NAVA) points performers to statutes like California's AB2602 and New York's S7676B. In Lehrman, the court let right of publicity claims proceed under New York's "digital replica" provision, which can apply to AI-generated voice clones even without any visual likeness.

Next, protect yourself before you ever step into a booth. NAVA recommends using an AI rider before recording sessions that covers consent, explicit limits on AI training use, term limits (beware contracts running "in perpetuity throughout the known universe"), payment, exclusivity, and storage controls. Without a rider, you may be signing away rights you didn't realize you had.

Finally, preserve evidence of how your voice was procured. In the Lovo case, the voice actors were allegedly hired anonymously through Fiverr under a "research purposes only" pretext, then their voices were sold commercially on Lovo's Genny platform under pseudonyms like "Kyle Snow" and "Sally Coleman." That deceptive procurement pattern helped support surviving contract and deceptive-practices claims, even though the plaintiffs were originally paid modest sums (Lehrman received $1,200 and Sage received $800). Save every message, job posting, and licensing term you can find.

• Check your state's Right of Publicity statute for voice and "digital replica" protections.
• Use a NAVA-recommended AI rider before every recording session.
• Preserve all communications showing how your recordings were obtained and used.
• If your recordings were used as training data, ask an attorney about biometric privacy claims like Illinois BIPA, which carries statutory damages of $1,000 to $5,000 per violation.

The same documentation discipline applies on the business side. If you're deploying voice AI, whether a narration tool or something customer-facing like an AI receptionist, vet how the provider sourced its voices. Look for consent-based licensing, transparent documentation, and security certifications such as SOC 2 Type I/II or ISO 27001, since gaps in licensing or data handling increase your legal exposure. Book a free setup call if you want to review how consent-based voice AI works in practice before you commit.

For Business Owners: Choosing an AI Voice Tool That Won't Create Legal Exposure

If your vendor cloned a voice without consent, the lawsuit may land on your desk, not theirs. That's the uncomfortable lesson of the current wave of voice litigation, and it applies just as much to the small business deploying an AI receptionist as to the tech giants in the courtroom.

The numbers explain why. Illinois' Biometric Information Privacy Act carries statutory penalties of $1,000 per negligent violation and $5,000 per intentional one, applied per victim, according to Reuters reporting on pending class actions against nine major tech companies. Plaintiffs' counsel in those cases points to prior BIPA settlements of $75 million (BNSF) and $47.5 million (Motorola Solutions) as a preview of the stakes. And the Lehrman v. Lovo ruling in July 2025 showed that even when federal copyright claims fail, state publicity, contract, and consumer-protection claims survive, as Fredrikson & Byron's analysis notes.

There's a second exposure layer: what your AI receptionist hears on calls. An AI phone tool that captures customer names, healthcare instructions, or payment details is handling data that GDPR and HIPAA regulate. Compliance guidance on AI voice risk is blunt: storing or reusing sensitive inputs without safeguards, or lacking licensing documentation and retention policies, increases legal exposure for the deploying business.

So how do you vet a provider before you sign? Ask these questions:

• Consent-based voice licensing. Can the vendor show exclusive, compensated contracts with the actual voice actors? The Lovo case involved voices procured under false "research only" pretexts and sold commercially, a pattern that kept contract and deceptive-practices claims alive.
• Security certifications. Look for SOC 2 Type I/II and ISO 27001. Their absence isn't proof of sloppiness, but their presence signals audited controls.
• A written policy against training models on your customer data, plus clear retention and deletion terms for call recordings and transcripts.
• Transparent disclosure that callers are speaking with AI, which consumer-protection statutes increasingly expect.

The good news is that reputable vendors make this easy to verify. Answrr, for example, deploys its AI receptionist with white-glove setup, so the compliance questions get answered during implementation rather than after a complaint arrives. If you're evaluating any voice AI tool for your practice or service business, book a free setup call and ask directly how voices are licensed and where call data goes.

The pattern in the Lovo case is the one to avoid: voices hired anonymously, used commercially under pseudonyms, with payments as small as $1,200 for work that ended up in court. A vendor who can't document consent is a vendor who's transferring that risk to you.