AI must not imitate other people’s voices
AI must not clone other people’s voices
Cloning the voice of Adele, Eminem or Karel Gott and distributing the recording? Technically, that is easy – but legally it is very dangerous! A voice is part of a person’s personality and is protected by several layers of law.
Modern AI tools can clone a voice from just a few seconds of audio. The result can be indistinguishable from the original. But behind every voice there is a real person – and that person has the right to decide how their voice is used. This article explains why imitating other people’s voices with AI is prohibited, which laws protect voices and what you risk if you break this rule.
The voice as part of personality
A voice is not just a sound – it is an identity
A voice is as unique as a fingerprint. Legal systems around the world therefore protect it as part of personality rights. This applies to:
Which laws protect it?
Protection of personality
The Civil Code (Act No. 89/2012 Coll., Sections 81–90) protects a person’s natural rights, including their likeness, expressions and voice. No one may capture, disseminate or imitate another person’s expressions without their consent in a way that could damage their honour or dignity.
§ 81–90 of Act No. 89/2012 Coll.Performers’ rights
The Copyright Act (Act No. 121/2000 Coll.) protects performers – singers, actors, musicians. Their performance (including their singing voice) must not be reproduced, adapted or disseminated without their consent. AI voice cloning directly infringes these rights.
§ 67–72 of Act No. 121/2000 Coll.GDPR and biometric data
A voice is biometric data within the meaning of the GDPR. Processing it (including training an AI model on someone else’s voice) requires the explicit consent of the person concerned. A breach of the GDPR can lead to fines of up to EUR 20 million or 4% of annual turnover.
Regulation (EU) 2016/679 (GDPR)EU AI Act
The Artificial Intelligence Act (in force since 2024, fully applicable from 2026) explicitly regulates systems for voice manipulation and deepfake content. It requires AI content to be labelled transparently and prohibits the deceptive use of synthetic voices.
Regulation (EU) 2024/1689GDPR and the voice – what does it mean in practice?
A voice is biometric data – a special category of personal data under Article 9 of the GDPR. For anyone to legally:
Specific scenarios
Prohibited: An AI cover with someone else’s voice
You generate a song “sung” by Beyoncé, Taylor Swift or Lucie Bílá – even if it is your own song. The voice is protected regardless of what it sings.
Prohibited: A deepfake recording
You create a recording in which a specific person “says” or “sings” something they never said. Even without distribution, this can be a criminal offence (defamation, damage to reputation).
Grey area: An “inspired” voice
You use AI to generate a voice that “sounds like” a particular artist but is not a direct clone. A legal dispute is still possible – especially if the intent to imitate is obvious from the marketing or the metadata.
Allowed: Your own AI voice
You create or license an AI voice model of your own voice and distribute recordings that use it. Or you use an AI voice from a tool licensed for that purpose (e.g. Suno or Udio with their own models).
Overview: What you can and cannot do
| Situation | Allowed? | Why |
|---|---|---|
| A famous singer’s AI voice on your own song | Prohibited | Infringement of performers’ rights + protection of personality |
| An AI clone of someone’s voice without their consent | Prohibited | GDPR – biometric data, protection of personality (Civil Code) |
| Distributing a deepfake recording | Prohibited | Breach of the Copyright Act and the Civil Code, potentially criminal law |
| An AI voice “inspired” by a specific artist | Risky | Depends on the degree of similarity and the intent – a dispute is possible |
| Your own AI voice (your own model) | Allowed | You hold the rights to your own voice |
| An AI voice from a licensed tool (its own models) | Allowed | The tool’s licence covers the use of its own voices |
How do platforms handle it?
Spotify, Apple Music, YouTube and the other major platforms explicitly prohibit in their terms the uploading of content that imitates a real person’s voice without their consent. Detection is both automatic and manual – based on reports from the artists themselves or their legal representatives. The result is immediate removal, demonetisation or a permanent account ban.
- Removal of the recording from all platforms without compensation.
- Permanent suspension of your distribution account (including all your other recordings) – on the platforms and with us.
- A civil lawsuit for infringement of personality rights – damages, an apology, a ban on dissemination.
- A GDPR fine – up to EUR 20 million or 4% of annual turnover.
- A criminal complaint – in the case of deliberate defamation or fraud.
- Penalties under the EU AI Act – for the deceptive use of a synthetic voice that is not labelled.
If you distribute a recording with an AI-generated voice (even your own or a licensed one), platforms still require AI content to be flagged in the metadata. Without this label, you breach the platforms’ terms and, from 2026, the EU AI Act as well. With us, you flag AI by answering YES to the question Is it AI-generated music? in the Track origin wizard – this also applies to an AI model of your own voice. TikTok, YouTube Content ID, Facebook Fingerprinting and Beatport are then not available for an account with AI-generated music.
You can read more about platforms automatically labelling AI music in the article Platforms are starting to label AI music.
The international dimension
In the United States, the voice and likeness of celebrities are protected by the so-called “right of publicity”. In many states (California, New York, Tennessee), this right is very strong and continues after death. In 2024, Tennessee passed the ELVIS Act (Ensuring Likeness Voice and Image Security), which explicitly protects artists’ voices from AI cloning.
Similar protection exists in other countries through a combination of copyright, personality rights and emerging AI legislation. Distribution on global platforms means that you are subject to the laws of every country where the content is available.