Is AI Voice Cloning Legal? Consent, Disclosure and the EU AI Act

AI voice cloning is legal in many situations - and clearly illegal in others. The deciding factors are almost always the same: whose voice it is, whether they agreed, and whether listeners could be misled. This article gives an overview of the rules that matter for creators and businesses.

The short answer

• Your own voice: generally fine to clone and use.
• Someone else's voice with their informed, written permission: generally fine, within the scope of what they agreed to.
• Someone else's voice without permission: risky to unlawful, and clearly unlawful when used to deceive, defraud, harass or imply an endorsement.

Why consent matters so much

A voice identifies a person as clearly as a face. Several areas of law can protect it at the same time:

Personality and publicity rights

Many jurisdictions protect a person's name, image and voice against unauthorised commercial use. Using a sound-alike or a clone of a recognisable person to sell a product can create liability even without any recording of theirs being copied. Some places have added explicit voice protections: Tennessee's ELVIS Act, for example, specifically covers unauthorised AI imitation of a person's voice.

Data protection

Under the GDPR, a voice recording of an identifiable person is personal data. When voice data is technically processed to uniquely identify someone, it becomes biometric data - a special category with stricter rules. Cloning another person's voice therefore needs a lawful basis, which in practice usually means their explicit consent.

Fraud, defamation and consumer protection

Using a cloned voice to trick someone, put words in a real person's mouth or fake an endorsement can fall under fraud, defamation or advertising law. Regulators are also acting on specific harms: in the United States, the FCC ruled in 2024 that AI-generated voices in robocalls count as "artificial" voices under existing telephone consumer protection rules.

The EU AI Act: transparency for synthetic media

The EU AI Act introduces transparency obligations for AI-generated content, which apply from 2 August 2026:

• Providers of AI systems that generate synthetic audio, images, video or text must ensure the outputs are marked in a machine-readable format and detectable as artificially generated.
• Deployers - the people and organisations who use such systems - must disclose when content is a deepfake: image, audio or video that resembles real people, places or events and could falsely appear authentic. Lighter rules apply to evidently artistic, satirical or fictional works, but the existence of AI-generated content must still be disclosed in an appropriate way.

In practice: if you publish a realistic AI voice that could be mistaken for a real person's recording, tell your audience clearly - in the content itself, not hidden in the terms.

A practical consent checklist

If you clone a voice that is not your own - a presenter, a client, an actor, a colleague - get written authorisation that covers:

1. Who may create and use the clone.
2. What for: the specific projects, channels and markets.
3. How long: the duration, and what happens when it ends.
4. Disclosure: how the audience will be told the voice is synthetic.
5. Tools and providers involved, and how the samples are stored.
6. Withdrawal: how the person can revoke consent and have the voice deleted.
7. Payment, if the voice is used commercially.

Keep the signed document. Pynokio may ask for evidence when it reviews a complaint.

What is never acceptable

• Cloning public figures, colleagues, customers, family members or deceased people whose rights are still protected without authority.
• Fraud, social engineering, bypassing voice authentication, fake evidence, political manipulation or fake endorsements.
• Sexual, humiliating, threatening or discriminatory content, and targeted harassment.
• Cloning a minor's voice through a standard self-service flow.

How Pynokio builds this in

• Consent before cloning. Creating a clone requires a declaration that the voice is yours or that you hold written authority. The declaration is recorded with a versioned statement and timestamp.
• Private samples. Voice samples are never served as public files.
• Watermark and provenance. Every generated voice carries an inaudible watermark and a provenance record, and anyone can check a file with the free Audio Detector. See how audio watermarking works.
• Deletion and complaints. Deleting a voice removes its samples and revokes its consent record. A voice owner who finds their voice misused can contact us, as described in the Voice Policy.

FAQ

Is it legal to clone my own voice?

Generally yes. You can clone and use your own voice, but you should still disclose synthetic speech where listeners could be misled.

Can I clone a celebrity's voice for a parody?

Not on Pynokio without the person's authority. Parody exceptions vary by country and rarely cover realistic voice clones; the risk of misleading listeners and infringing personality rights is high.

Do I have to say that my content uses an AI voice?

If the voice could be mistaken for a real person's recording, yes - the EU AI Act requires disclosure of deepfakes, and many platforms require labels for realistic synthetic media. Disclosing is good practice in every case.

Is my voice sample safe on Pynokio?

Samples are private, are not served as public files, and are deleted when you delete the voice.

What should I do if someone cloned my voice without permission?

Collect the evidence, check the file with the Audio Detector, and contact the platform where it appeared. If it was made with Pynokio, contact us as described in the Voice Policy.