Cloning your own voice is fine. Cloning someone else's without written permission is now illegal in a growing list of places, and the penalties are real. Tennessee's ELVIS Act, California, New York, and others require consent before you copy a person's voice, and the EU treats a voiceprint as biometric data. This guide covers whose permission you need, what the 2026 laws say, and how to use voice cloning without a legal problem. It is general information, not legal advice, so check with a lawyer for your own case.
Voice sits inside the wider disclosure rules; our guide to AI content compliance and the EU AI Act covers the labeling side that pairs with the consent rules here.
Is AI voice cloning legal?
It depends on whose voice you clone. Your own voice, or a fully synthetic voice the tool provides, is safe to use within the tool's terms. The moment you clone a real other person, consent becomes the line between legal and not. Copying a public figure or any identifiable person without permission is what the new laws target, whether the result is a parody or an ad.
The tool does not hand you the right to the voice. A paid ElevenLabs plan, for example, grants commercial rights to the output, but its terms still require that you own or have permission for the voice you cloned. Paying for the software is not the same as clearing the voice, and creators trip on that difference constantly.
Getting it wrong is expensive. The newer state laws let the person sue directly, often with enhanced damages for intentional misuse, and platforms increasingly must take the clip down on request. A viral prank that clones a celebrity, or an ad that fakes a well-known voice, is exactly what these statutes were written to punish. Reach does not reduce the risk here; it raises it.
The voice-cloning laws you have to know
The rules tightened fast, led by the states rather than Washington:
- Tennessee ELVIS Act (2024). The first law aimed squarely at AI voice mimicry; it bans cloning a person's voice without permission and lets that person sue.
- California and New York. California Civil Code 3344 and New York's Civil Rights Law 50-51 protect voice and likeness, and both states have moved to sharpen them for AI.
- A widening list. Texas, Illinois, and more have introduced or toughened voice-cloning statutes, many adding a private right of action and platform takedown duties.
- Federal (NO FAKES Act). A national right-of-publicity bill for AI clones has not passed as of 2026, so the state laws are what bind you today.
The pattern across them is consistent: an identifiable person owns their voice, and copying it without a yes is now something they can take to court.
One question decides most cases: is the voice identifiable? The laws protect a voice that listeners would recognize as a specific person, so a soundalike that clearly evokes a known individual can trip the same wire as an exact clone. A generic synthetic voice that resembles no one in particular sits outside that, which is why licensed stock voices stay the low-risk default for commercial work.
Federal agencies moved too, even without a new statute. The FTC has acted against AI voice used in scams and impersonation, so cloning a voice to deceive, such as faking a company or a person to mislead, draws consumer-protection liability on top of the state right-of-publicity laws. Two separate legal tracks can apply to the same clip.
Do you need consent, and what kind?
For anyone else's voice, yes, and a casual "sure, go ahead" does not hold up. The state laws point to written consent that names the specific use and is documented so you can prove it, and that stays revocable where the law requires. A signed release naming the project and its scope is the safe standard, the same way a model release works for a face.
Your own voice leaves you clear, though it still pays to keep the consent record if a client or platform asks. For a hired voice actor, the release has to cover AI cloning explicitly, because an old contract for a normal recording almost never granted the right to synthesize new lines from it. That gap is exactly where many recent disputes began.
Two edge cases catch people out. When you commission a clone from a contractor, put in writing who owns the resulting voice model, or you may not control the thing you paid for. And cloning a minor's voice raises the bar, since consent has to come from a parent or guardian, and several statutes treat misuse involving minors more harshly. When in doubt, get the release in writing and keep it on file.
The EU treats a voiceprint as biometric data
Europe adds a second layer on top of consent. Under the EU AI Act, in full effect in 2026, synthetic audio that could pass for a real person must be labeled as AI-generated, which our guide on labeling AI content walks through step by step. Beyond labeling, a voiceprint counts as biometric data, so processing the voice of an EU resident needs a lawful basis under GDPR, usually explicit consent.
In practice that means two boxes for EU-facing work: a lawful basis to make the clone, and a perceptible disclosure when you publish it. Skipping either is its own risk, and the fines under the AI Act reach into the millions of euros.
The GDPR side is easy to overlook, because it applies even to a voice you have permission to use. Explicit consent is the cleanest lawful basis, but it has to be freely given and specific to voice processing, not buried in a general terms checkbox. For EU residents you also owe them a way to withdraw that consent later, which in practice means keeping the source recording and the clone tied to a record you can actually act on.
How do you clone voices without legal risk?
Keep it simple and you stay clear:
- Clone your own voice, or use a tool's licensed stock voice, whenever you can.
- For anyone else, get written, specific consent before you clone, and keep the document.
- Make sure a voice actor's release names AI cloning explicitly, not just the original recording.
- Disclose the AI voice where required, especially for EU audiences and on platforms that ask.
- Read your tool's terms: commercial rights to the output still assume you cleared the voice.
The single most common way creators get burned is treating a famous voice as fair game for a joke. Satire has some protection in places, but a cloned celebrity voice in anything that looks commercial is the highest-risk move you can make, and it is the fact pattern behind most takedowns and suits. A simple test: if you would not use the person's actual recording without a deal, do not clone their voice either.
Do that and voice cloning is a normal production tool rather than a lawsuit waiting to happen. For the platform side, our guide on disclosing an AI voice without losing monetization covers where the toggles live. And once more, this is general guidance, not legal advice for your situation.
Want to use AI voice and video the right way, professionally? The Future Tech program teaches AI content production, including the compliance habits that keep your work publishable.
