AI Voice Cloning: Copyright & Right of Publicity Risk for Businesses 2026
Businesses cloning a voice for an AI customer-service line, an ad, or a podcast intro often ask the wrong legal question first: "who owns the copyright in this voice?" The honest answer is nobody — voices aren't copyrightable. The real exposure sits in a different body of law that most AI-copyright checklists skip entirely.
Copyright Doesn't Cover Voices — Right of Publicity Does
Copyright protects fixed, creative expression: a specific recording, a script, a musical arrangement. A person's voice — its timbre, cadence, and acoustic fingerprint — isn't a "work" in the copyright sense, so there's no copyright to infringe by training a model on it or generating new speech in that voice. This is a genuinely different legal regime than the copyright exposure covered elsewhere in AI content generation, and businesses that treat voice cloning as a copyright problem end up checking the wrong box entirely.
The actual exposure runs through right of publicity — a state-law doctrine that protects a person's name, likeness, and, increasingly, voice from unauthorized commercial use. It doesn't require the voice to be "original" or "creative" the way copyright does; it only requires that the voice is identifiable as belonging to a specific person and that it was used commercially without consent.
Tennessee's ELVIS Act: The Template Other States Are Watching
Tennessee's Ensuring Likeness Voice and Image Security Act amended the state's existing right-of-publicity statute to explicitly add "voice" as a protected attribute and to address AI-generated replicas directly. It creates liability not just for using an unauthorized voice clone commercially, but for distributing or making available a tool whose primary purpose is producing unauthorized clones of a specific person's voice — reaching software vendors, not just the businesses that use the output. Other states with strong entertainment-industry lobbying, including California, are expected to follow with similar voice-specific amendments.
The Midler Precedent Still Matters
Long before AI voice cloning existed, the Ninth Circuit's decision in Midler v. Ford Motor Co. established that imitating a distinctive, identifiable voice for a commercial — even without copying an actual recording — can violate right of publicity if consumers would recognize whose voice it's meant to evoke. AI voice cloning tools make this dramatically easier and more precise than a human sound-alike ever was, which is exactly why courts and legislators are treating it as the same underlying harm at a much larger scale.
Where Businesses Actually Get Exposed
- Cloning a voice actor's voice from prior session recordings to generate new lines without a new contract covering AI use
- Using an AI-cloned celebrity or influencer voice in an ad without a publicity license, even if no copyrighted recording was copied
- Deploying an AI customer-service voice modeled closely on a real employee's voice after that employee leaves the company
- Using a voice-cloning tool built for a different purpose (e.g., dubbing) to recreate a specific real person's voice for marketing
- Assuming a "royalty-free voice library" license covers AI training and cloning rather than just playback of the original recordings
The NO FAKES Act: Federal Law Is Coming, Not Here Yet
The federal NO FAKES Act would create a nationwide, IP-like right against unauthorized AI-generated voice and visual likeness replicas, with statutory damages and safe harbors for platforms that remove infringing content. It has bipartisan support and backing from major entertainment and music industry groups, but has not been enacted. Until it is, businesses face a patchwork where the same voice-cloning use case might be low-risk in one state and high-risk in another, based purely on which state's right-of-publicity law applies.
Compliance Checklist
Treat voice cloning as a consent and licensing problem, not a copyright-clearance problem.
Voice licensing doesn't fix a site users can't navigate
Clearing your AI voice content is a legal issue separate from whether people using screen readers or keyboard navigation can actually use your site. RatedWithAI scans your site for the accessibility issues that turn into complaints and lawsuits.
Scan Your Site for Free →Frequently Asked Questions
Can a business copyright the output of its own AI voice model?
The synthetic audio output may qualify for copyright protection as a sound recording if there's sufficient human creative input in producing it, but that copyright — if it exists — belongs to whoever made the recording, not to the person whose voice was cloned. It also doesn't shield the business from a separate right-of-publicity claim by the person whose voice was used without consent.
Does using a text-to-speech voice from a commercial AI vendor avoid this risk?
Generally yes, if the vendor's stock voices are synthetic composites not modeled on a specific identifiable person, or if the vendor has properly licensed the voice talent behind them. The risk arises specifically when a business clones a particular real person's voice, whether through a vendor's cloning feature or an open-source tool.
What damages are available in a right-of-publicity voice cloning claim?
Remedies vary by state but commonly include actual damages, disgorgement of profits attributable to the unauthorized use, and in some states statutory damages or injunctive relief barring further use of the clone. States with explicit AI-voice provisions like Tennessee's ELVIS Act have added remedies aimed specifically at cloning tools and their distributors.
Does consent from a voice actor need to be renewed for each new AI-generated project?
It depends on how the original consent was scoped. A narrowly scoped release covering only a specific project or timeframe will not automatically extend to new, unrelated commercial uses of the AI-cloned voice, which is why broad, AI-specific licensing language matters more than a one-time signature.