AI Voice Cloning Consent Rules: Whose Voice You May Use

The rule is short: clone your own voice freely, clone someone else's only with their documented, specific permission, and never clone a real identifiable person without it. The fact that the software can copy any voice from a few seconds of audio is not permission, and audio being publicly available online is not permission either. Those two misconceptions cause most of the trouble. Everything else is detail about how to document consent and how it differs from a license.

This article describes what the platform policies and the law actually say, read from primary sources on 2026-09-11, and routes you to them. It is not legal advice. Voice-likeness law is changing quickly and varies by where you and your subject are, so treat this as a map of the questions to ask, and take the answers that carry real risk to your own counsel.

AI voice cloning consent: the three cases, in order of safety

Almost every real situation is one of three, and naming which one you are in settles most of the decision.

Your own voice. This is the safe base case. Cloning your own voice to narrate your own content raises no consent problem, because you are the person who can grant permission and you are granting it to yourself. The open questions here are practical, about quality and about protecting the model from misuse, not about whether you may.

Someone else's voice, with permission. This is allowed when you have that person's genuine, informed, specific and documented consent, plus, for most commercial uses, a license that spells out where and how you may use the result. A quick verbal yes or a buried checkbox is thin evidence for anything commercial. Get it in writing, name the uses, and keep it.

A real identifiable person, without permission. Do not. This is the case that generates lawsuits, platform bans and, increasingly, statutory liability. It does not matter that the audio was easy to find or that the clone is "just for a video." Absent consent or a clear legal right, this is the case to walk away from.

Consent and a license are two different questions

The single most useful distinction here, and the one the quick guides tend to blur, is that consent and licensing answer different things.

Consent answers whether you are allowed to create the clone at all. A license answers how, where, for how long and for what purpose you are allowed to use it once it exists. You can have one without the other, and both usually matter. A voice actor might consent to being cloned for one campaign and not for another; the license is where that boundary lives. So "they said yes" is not the end of the conversation. Yes to what, for how long, on which platforms, and can they revoke it, are the questions a real agreement answers. Treating consent as a single yes or no is how well-meaning creators end up using a voice far beyond what its owner actually agreed to.

What the platforms already require of you

The tools themselves encode a version of these rules, and their policies bind you the moment you sign up, before any court is involved.

ElevenLabs is a clear example. Its Prohibited Use Policy, Section 5 (read at elevenlabs.io/use-policy on 2026-09-11), states that you must not create audio that intentionally replicates the voice of another person "without consent or legal right," nor in a way that harasses or harms that person, nor in a manner intended to deceive others about whether the voice was AI-generated. Separately, its safety page (read the same day) describes "blocking the cloning of celebrity and other high risk voices, and requiring technological verification for access to our Professional Voice Cloning tool." In other words, the more capable cloning tools gate access behind a verification step precisely because consent is the risk they are managing.

The practical reading: the platform is not going to take your word for it on high-risk voices, and its terms already put the consent obligation on you. Violating that policy can cost you the account regardless of what any statute says.

The legal layer, named plainly and routed to counsel

Beyond the platforms, two kinds of law are in play, and this is the point to be careful and general rather than specific.

The first is the right of publicity, the long-standing principle that a person controls the commercial use of their identity, which many US states recognize and which courts have applied to imitations of a distinctive voice. Several states have recently gone further and passed laws aimed specifically at unauthorized AI voice and likeness replicas. Because the number of states, the exact scope and the remedies differ and are moving, this article does not pin a specific count or statute to your situation. What it tells you is that a recognizable voice is a protected interest in many places, and cloning one commercially without consent can create liability under that body of law.

The second is impersonation and fraud enforcement. The US Federal Trade Commission has publicly moved to strengthen protections against AI impersonation of individuals, warning that "fraudsters are using AI tools to impersonate individuals with eerie precision and at a much wider scale" and that voice cloning is central to the rise in impersonation scams (ftc.gov press release, read 2026-09-11). That is an enforcement posture, not a consent framework, but it tells you which uses draw regulators: anything that deceives someone about who is speaking.

Neither of these is something a creator should self-diagnose on a live project. Where a voice-cloning use carries real commercial or reputational stakes, that is the moment for your own lawyer and the actual statute in your jurisdiction, not a blog.

Two edge cases worth flagging

Some situations feel like exceptions and mostly are not.

A voice that "just sounds like" a celebrity, produced deliberately to evoke them, is not a safe loophole. The right of publicity has reached imitations, not only literal copies, when the intent is to trade on the person's identity. And a deceased person's voice is not automatically free to use; several states extend publicity rights after death, and the estate may control them. If a monetized channel is built on a synthetic persona, note too that platform originality and category rules apply on top of all of this, which this site covers in the piece on the platform's inauthentic-content monetization rules.

FAQ

Can I legally clone my own voice? Yes. Cloning your own voice for your own use raises no consent issue, because you are the person entitled to grant permission. The remaining questions are practical ones about quality and about keeping the model from being misused, not about whether you are allowed.

Is a voice legal to clone if it is publicly available online? No. Public availability is not permission. As the platform policies and right-of-publicity principles make clear, you generally still need the person's consent, and for commercial use a license, regardless of how easy the source audio was to find.

What counts as proper consent? For anything commercial, consent should be informed, specific, documented and tied to a license that names the permitted uses. A verbal yes or a buried checkbox is weak evidence. Put it in writing, state what the voice may be used for and for how long, and keep the record.

What is the difference between consent and a license? Consent is permission to create the clone at all. A license governs how, where and for what you may use it once created. You typically need both, and a yes to being cloned is not a yes to every possible use.

Where to check before you act

Start with your tool's own use policy, which already puts the consent duty on you and can cost you the account if you ignore it. For ElevenLabs that is elevenlabs.io/use-policy, read 2026-09-11. For the enforcement climate, the FTC's public statements on AI impersonation are the primary source. For the law that would actually apply to your project, read the statute in your jurisdiction and talk to your own counsel. The one rule that never changes: capability is not consent.