Your friend sends you a voice note. You run it through an AI tool, and minutes later, you send back a version of their voice announcing their intention to become a professional pickleball player.
Funny? Sure. Harmless? Maybe. But once the clip leaves the group chat and starts fooling people, the legal questions get more serious. AI voice cloning can blur the line between a harmless joke, an embarrassing invasion of privacy, and potential impersonation or fraud.
The safest rule is simple: get permission before cloning anyone’s voice, and be clear about how you plan to use it. But there is more to consider before cloning your friend’s voice.
Why Voice Cloning Gets Messy
AI voice cloning tools can turn a short video, voice note, podcast clip, or social-media post into a convincing imitation. The U.S. Copyright Office describes a digital replica as audio, video, or imagery digitally created or manipulated to realistically but falsely depict a person.
That can start as harmless fun, but it becomes risky when people think the audio is real. The Federal Trade Commission warns that scammers can use cloned voices to impersonate relatives, executives, and other trusted people, often to invent emergencies or request money. The FTC advises people to independently verify an urgent request using a phone number they already know belongs to the person.
Money is not the only issue. A fake clip can embarrass someone, harm their reputation, invade their privacy, or make them appear to say or endorse something they never did. Before you clone your friend’s voice, ask whether they would be comfortable hearing the result played to people who might believe it is real.
The Law Is Still Catching Up
No single nationwide law governs every situation involving the AI cloning of another person’s voice. Existing laws involving fraud, impersonation, harassment, defamation, privacy, consumer protection, copyright, and publicity rights may all come into play depending on what you make, where you share it, and what harm results.
States are beginning to enact laws aimed at AI-generated replicas, but their scope varies. Tennessee’s ELVIS Act, effective July 1, 2024, expanded the state’s personal-rights protections to include an identifiable simulation of a person’s voice. It creates civil liability for certain knowing, unauthorized public uses and includes First Amendment protections for qualifying uses such as news, comment, criticism, scholarship, satire, and parody.
California’s AB 2602 is narrower. It regulates certain personal- and professional-services contracts involving digital replicas; it is not a general ban on consumer AI voice cloning. Still, it reflects a broader legislative push to require more specific consent when a person’s voice or likeness may be used in place of their work.
These laws do not resolve every private prank or online post. Depending on the facts, noncommercial voice cloning can still raise issues under laws governing fraud, defamation, harassment, privacy, or impersonation. The National Conference of State Legislatures’ AI legislation database tracks state proposals and enacted AI laws, while its deepfake-legislation overview shows lawmakers’ growing focus on deceptive synthetic audio and video.
Consent Is Key
A friend’s permission can make a real difference, but it needs to be specific. Agreeing to a clearly labeled AI joke for a private birthday video or short voiceover is not the same as agreeing to have their voice posted publicly, used in an ad, turned into a song, made to say something sexual or inflammatory, or uploaded to a platform that keeps the voice model or recordings.
Before cloning your friend’s voice, agree on what the audio will say, where it will appear, who will hear it, whether it will be labeled as AI-generated, and whether the recording or voice model will be deleted afterward. One-time permission does not authorize every future use.
Then ask the more important question: would a reasonable person think the audio is real? If it makes your friend appear to say or endorse something they did not, could harm their reputation, relationships, job, finances, or safety, or is meant to impersonate them or influence someone, do not share it. A disclaimer may reduce confusion, but it does not turn a deceptive or harmful voice clone into a good idea.
Keep It Recognizable as a Joke
A clearly labeled parody shared privately is very different from an AI-generated recording intended to convince a friend’s employer that they quit.
The first may still require permission, especially if it is public. The second can quickly raise issues far beyond awkward group-chat energy. The same goes for cloning someone’s voice to contact their bank, partner, employer, customers, or relatives. That is not a prank with good production value; it is potentially impersonation or fraud.
And if the “friend” is a celebrity, creator, musician, coworker, former partner, or someone you do not actually know, assume the risk goes up, not down.
The Bottom Line
Clone your friend’s voice only with their clear permission, for a defined and nondeceptive purpose, and label the result as AI-generated when anyone else might hear it. If the bit depends on people believing your friend actually said it, the joke is probably not worth the fallout.
