Civil Law and Voice Cloning Unauthorized Use Disputes in Europe

1. Introduction

Voice cloning is the creation of a synthetic or digitally manipulated voice that reproduces the distinctive vocal characteristics of a real person. Modern AI systems can reproduce:

tone;

pitch;

accent;

rhythm;

pronunciation;

vocal timbre;

speech patterns;

characteristic expressions.

This creates a significant civil-law problem when a person's voice is reproduced without consent, particularly where the cloned voice is used:

in advertising;

in films or entertainment;

in political content;

on YouTube or social media;

for commercial endorsements;

in audiobooks;

in music;

in pornography or defamatory material;

in fraudulent communications;

to impersonate a celebrity;

to train or operate an AI voice system.

European law does not yet have one single harmonised "right to one's voice" statute. Instead, protection comes from several overlapping legal doctrines:

personality rights;

privacy rights;

image/likeness protection by analogy;

name and publicity rights;

tort/delict law;

unjust enrichment;

copyright and related rights in particular recordings;

data protection law;

consumer protection;

contractual law;

freedom of expression and artistic freedom.

Importantly, the case law is evolving rapidly. A particularly important development occurred in 2025–2026, when German and French courts directly addressed AI-generated or reproduced voices.

2. The Central Civil-Law Question

The central question is:

Does an individual have a legally protected interest in controlling the commercial and personal exploitation of their identifiable voice, including an AI-generated imitation of that voice?

The emerging European answer is increasingly yes, although the precise legal basis and remedies differ between countries.

A voice does not necessarily need to be an original copyrighted work.

Instead, the claimant may argue:

"That voice identifies me as a person, and the defendant has appropriated that identifying personality characteristic without my consent."

This is fundamentally a personality-rights claim.

3. The Most Important Modern Case: German AI Voice-Cloning Litigation

Case 1 — LG Berlin II, 20 August 2025, 2 O 202/24

Court: Landgericht Berlin II
Date: 20 August 2025
Subject: AI-generated imitation of a professional voice actor

This is one of the most directly relevant European authorities concerning AI voice cloning.

The claimant was a well-known German voice actor whose voice was used in connection with two YouTube videos.

The defendant used AI-generated audio closely resembling the claimant's distinctive voice.

The claimant had not consented to the use.

The videos did not adequately communicate to viewers that the voice was AI-generated, and the similarity was sufficiently strong that members of the public believed the claimant had actually participated in the videos.

Court's reasoning

The court recognised the individual's right to his own voice as part of the general personality right.

It considered an AI-generated imitation comparable, in principle, to imitation by a human voice impersonator.

The critical issue was not whether the defendant had literally copied an original sound recording.

Rather:

Was the claimant's recognisable vocal identity deliberately reproduced?

The court concluded that the use infringed the claimant's personality rights.

Commercial significance

The defendant used the voice in videos that generated commercial value, including increased attention and monetisation.

The court therefore recognised an economic dimension to the right.

A particularly important remedy was the possibility of a fictitious licence fee.

This means that the claimant does not necessarily have to prove:

"I lost exactly €X."

Instead, the court can potentially ask:

"What would the defendant reasonably have had to pay to lawfully obtain permission for this use?"

That is extremely important for AI voice-cloning disputes.

Principle

AI-generated imitation of a recognisable person's voice can constitute an infringement of personality rights even though the defendant did not use the person's original recording.

4. Case 2 — OGH, 6 Ob 287/02b, 20 March 2003 — MA 2412

Court: Austrian Supreme Court (Oberster Gerichtshof)

This is a landmark European voice-imitation decision predating generative AI.

The plaintiffs were actors in the Austrian television series MA 2412.

Their voices had distinctive characteristics, including:

tone;

pitch;

speech melody;

dialect;

manner of speaking.

Those characteristics were imitated in a political advertising broadcast.

Issue

Could a person object when their actual recording was not necessarily used but their recognisable voice was imitated?

The Austrian Supreme Court answered essentially yes.

It treated the identifiable voice as a protected personality characteristic.

The court considered unauthorized voice imitation in advertising to infringe personality rights and used the statutory protection of a person's likeness as an important analogy.

Why this case is crucial for AI

The technological method is different:

2003:

Human imitator → reproduces characteristic voice.

2025:

AI model → reproduces characteristic voice.

The legal question is substantially similar:

Is the person's identifiable vocal identity being exploited without permission?

Principle

Protection of personality can extend beyond an original recording to an identifiable imitation of the person's characteristic voice.

This is perhaps one of the strongest pre-AI European precedents for modern voice-cloning disputes.

5. Case 3 — OLG Hamburg, 3 W 45/89, 8 May 1989

Court: Hanseatic Higher Regional Court of Hamburg

This case involved the imitation of the voice of the famous German comedian Heinz Erhardt by a voice imitator.

The case is important because it demonstrates that German personality-right doctrine recognised the economic and personal significance of a distinctive voice decades before AI.

The court treated the intensity of the interference with a well-known voice as potentially comparable to the unauthorised use of:

a person's image;

a person's name;

another identifying personality characteristic.

Importance for voice cloning

AI cloning creates essentially the same legal problem at a higher technological level.

A human impersonator may need:

training;

acting skill;

recording equipment.

An AI system can create the same effect automatically.

But technological sophistication does not necessarily eliminate the underlying personality-rights issue.

Principle

A distinctive and recognisable voice can possess personality and commercial value sufficient to justify civil protection against unauthorised exploitation.

6. Case 4 — BGH, I ZR 49/97, 1 December 1999 — Marlene Dietrich

Court: German Federal Court of Justice (Bundesgerichtshof)

This is one of the most important German personality-rights cases.

The case concerned unauthorised commercial exploitation of Marlene Dietrich's personality, including her image and name.

Although it was not an AI voice-cloning case, the BGH expressly recognised that personality characteristics such as a person's:

image;

name;

voice;

other identifying characteristics

can have substantial economic value.

Commercial personality rights

The court recognised that famous individuals can commercially exploit their identity.

For example, a celebrity can authorise a company to use:

"my image + my name + my recognisable identity"

for advertising in exchange for payment.

If a third party does the same thing without authorisation, the economic value of the personality right may have been appropriated.

Application to AI voice cloning

Suppose an AI company creates:

"the voice of Celebrity X"

and sells it to advertisers.

The celebrity can argue:

"You have taken an economically valuable attribute of my personality and commercially exploited it without my consent."

That argument fits naturally within the reasoning of Marlene Dietrich.

Principle

Personality rights can have both ideal and economic dimensions, and unauthorised commercial exploitation can generate a damages claim.

7. Case 5 — BGH, VI ZR 104/57, 20 May 1958 — Tonbandaufnahmen

Court: German Federal Court of Justice

This is an older but foundational German personality-rights decision concerning unauthorised sound recording.

The court recognised that a person has significant control over whether their speech and voice are recorded and subsequently reproduced.

The court emphasised the importance of personal autonomy and the ability of a person to decide:

to whom their words are communicated;

whether their voice is recorded;

whether the recording is reproduced;

in what context it is subsequently heard.

Why it matters today

Voice cloning often begins with the collection of voice material.

For example:

Person gives an interview.

The defendant obtains the recording.

The defendant then:

extracts the voice → trains an AI model → creates new statements → publishes them.

The legal issue therefore may begin before the cloning itself.

There can be separate questions concerning:

acquisition of the recording;

retention of the recording;

extraction of voice characteristics;

creation of the clone;

publication of synthetic speech.

Principle

Control over one's recorded voice is part of the protection of personal autonomy and personality.

8. Case 6 — BGH, VI ZR 12/19, 26 November 2019

Court: German Federal Court of Justice

This case further developed German protection of the right of self-determination over spoken words and recordings.

The BGH explained that recording a person's voice does more than preserve the informational content of speech.

It also fixes:

vocal expression;

manner of speaking;

tone;

individual characteristics.

That creates a particularly strong form of personal appropriation.

Relevance to AI

An AI voice model may capture not merely:

"what the person said"

but:

how the person sounds.

This distinction is crucial.

A transcript saying:

"I agree with the proposal"

may not reproduce the person's voice.

An AI clone saying exactly the same sentence in the person's distinctive voice creates a much stronger personality-rights issue.

Principle

The law may protect not merely the informational content of speech but the individual's control over the vocal expression through which that speech is communicated.

9. Case 7 — Cour de cassation, First Civil Chamber, 24 June 2026, No. 25-20.483

Court: French Court of Cassation
Date: 24 June 2026

This is an exceptionally important current European authority.

The dispute concerned the reproduction of a person's voice in the song "Les Gens beaux."

The defendant argued, among other things, that the voice and statements had been publicly available and that artistic freedom justified the reproduction.

The Court of Cassation's major statement

The French Supreme Court expressly held that:

a person's voice is one of the principal attributes of personality.

It placed the right to respect for one's voice within the protection associated with:

Article 9 of the French Civil Code;

Article 8 ECHR.

The court further explained that the right to voice must be balanced against:

freedom of expression;

artistic freedom.

This is particularly significant

The court did not say:

"Every reproduction of a voice is automatically unlawful."

Instead, it required a balancing exercise.

Relevant factors include:

whether there is a public-interest debate;

the nature of the work;

the person's public status;

the circumstances in which the voice was originally used;

the artistic purpose;

the effects of the use.

Why this matters for AI

Imagine an AI-generated voice used in:

A commercial advertisement

versus

A political satire

versus

A documentary

versus

a fraudulent investment advertisement.

The legal outcome may differ substantially.

Principle

The voice is an independently protected personality attribute, but protection must be balanced against freedom of expression and artistic freedom.

This is probably the most important recent French authority for future European voice-cloning litigation.

10. Case 8 — Cour de cassation, First Civil Chamber, 6 December 2017, No. 16-21.679

Court: French Court of Cassation

This case involved unauthorised recording and broadcasting of a person's image and voice in a television programme.

The claimant had been recorded through a hidden camera.

The court considered whether the claimant was identifiable, including through:

body characteristics;

manner of expression;

intonation;

voice.

Importance

The case illustrates that the voice can contribute independently to the identification of a person.

This is important in voice-cloning disputes because the claimant does not necessarily need to prove:

"Everyone knew it was my voice."

The relevant question may instead be:

Could a sufficiently significant group of people identify the person from the vocal characteristics?

Principle

Voice and vocal mannerisms can contribute materially to personal identification and therefore to privacy/personality protection.

11. Case 9 — BGH, I ZR 2/21, 24 February 2022 — Tina Turner

Court: German Federal Court of Justice

This case concerned advertising for a Tina Turner tribute show.

The performer was not Tina Turner herself.

The issue concerned whether the use of a look-alike could create a misleading impression that the real celebrity was involved.

The BGH held that a depiction by another person can amount to a protected representation of the famous person when it creates a sufficiently convincing impression that viewers are seeing the person themselves.

Relevance to voice cloning

The reasoning can be adapted by analogy.

Consider:

"Tina Turner look-alike"

versus

"Tina Turner AI voice."

If a substantial portion of the audience believes that the real person:

participated;

endorsed;

authorised;

narrated;

the representation becomes legally more problematic.

Principle

The law can take account of audience perception and the likelihood that the public will believe that the real personality is involved.

That principle is highly relevant to AI-generated voice impersonation.

12. The Emerging European Rule

The cases collectively demonstrate an important evolution:

Traditional law

Voice = part of personal identity.

Commercial personality law

Voice = potentially economically valuable personality attribute.

Digital law

Digital reproduction of voice = potentially protected personality characteristic.

AI era

AI-generated imitation can infringe the right to one's voice even where the original recording was not copied.

This is a major doctrinal development.

13. What Exactly Is Protected?

A claimant may seek protection over several different things.

A. The actual recording

Example:

A recording of a singer's interview is copied.

Potential claims:

privacy;

copyright/related rights;

personality rights;

contractual claims.

B. The voice itself

Example:

A synthetic system generates new sentences in the singer's recognisable voice.

This is primarily a personality-right issue.

C. Voice style

Example:

A singer's distinctive vocal manner is deliberately replicated.

This is more difficult.

A general style is less easily monopolised than a clearly identifiable voice.

The stronger the identification, the stronger the claim is likely to become.

D. False attribution

Example:

AI voice says: "I recommend this investment."

The real person never said it.

The problem now involves:

personality;

deception;

reputation;

potentially defamation;

consumer protection;

fraud.

14. Identification Is Central

A voice-cloning claim generally becomes stronger when the public can identify the claimant.

Consider three situations.

Situation A

Generic male voice.

"Hello, welcome to our website."

Weak personality claim.

Situation B

Voice resembles a famous actor.

Public reasonably thinks the actor participated.

Much stronger.

Situation C

Voice is almost indistinguishable from the actor.

Advertisement implies the actor endorsed the product.

Very strong potential claim.

Thus:

Recognisability is one of the central factual questions in voice-cloning litigation.

15. Commercial Use

Commercial use substantially strengthens many claims.

Examples:

advertisements;

product endorsements;

audiobooks;

games;

films;

AI subscriptions;

paid social-media content;

virtual assistants;

political fundraising;

cryptocurrency promotions.

Suppose a company generates:

"Celebrity X's voice recommends Product Y."

If Celebrity X never agreed, potential claims may include:

personality-right infringement;

unfair competition;

passing off or equivalent national doctrines;

consumer deception;

unjust enrichment;

damages.

16. Non-Commercial Use

The situation becomes more complicated where the voice is used for:

parody;

satire;

political criticism;

documentary;

journalism;

artistic experimentation.

The French 2026 case is especially important here because it emphasises the need to balance:

personality rights

against

freedom of expression.

Therefore, there is no universal rule that:

"Any AI imitation is unlawful."

Context matters.

17. AI Act and Voice Cloning

As of 2 August 2026, the EU AI Act's transparency provisions concerning synthetic content are applicable.

Article 50 requires relevant AI systems producing synthetic audio, image, video or text to support machine-readable marking, and deployers of AI systems generating or manipulating audio, image or video constituting a deepfake must disclose that the content has been artificially generated or manipulated, subject to specified exceptions. (EUR-Lex)

The European Commission's 2026 guidance specifically recognises the problem of AI-generated or manipulated content that can create deception, impersonation and manipulation. (Digital Strategy)

Important distinction

The AI Act does not simply say:

"You need permission from the person whose voice is cloned."

Instead, it establishes transparency obligations.

Personality rights and civil liability therefore remain separately relevant.

Thus:

AI Act compliance ≠ permission to use someone's voice.

A company might correctly label:

"AI-generated voice"

and still face a personality-rights claim from the person whose voice was cloned.

18. GDPR and Voice Cloning

Voice data can also implicate the GDPR.

The European Data Protection Board has explained that voice data used for user identification can constitute biometric data under the GDPR when processed through technical means allowing unique identification or authentication. (European Data Protection Board)

Therefore, a voice-cloning project may involve two separate questions:

Personality law

"Did you unlawfully exploit my identity?"

Data protection law

"Did you unlawfully process my personal/biometric data?"

These claims can coexist.

19. Copyright and Voice Cloning

An important distinction must be made:

A person's voice is not automatically a copyrighted work.

Copyright generally protects qualifying creative expression.

However, a voice-cloning dispute may involve copyright where the defendant copies:

an original recording;

a musical performance;

an audiobook;

a protected vocal performance;

a protected sound recording.

Therefore, there may be:

Voice/personality claim

plus

copyright/related-right claim.

The two should not be confused.

20. Consent

Consent is one of the strongest defences.

A performer may sign:

"The producer may use my recorded voice for AI-generated synthetic performances."

The legal analysis then depends on the scope of the agreement.

Important contractual questions include:

Is AI cloning expressly permitted?

For how long?

For which territories?

For which products?

Can the voice be sublicensed?

Can the model be trained on the recordings?

Can the model generate new sentences?

Can the voice be used after termination?

Can it be used for advertising?

Can it be used for political content?

A vague clause such as:

"Producer may use recordings for promotional purposes"

may not necessarily answer all AI-cloning questions.

21. Training Versus Output

This distinction is particularly important.

Stage 1 — Training

The defendant collects 50 hours of recordings and uses them to train a voice model.

Stage 2 — Model creation

The model learns the speaker's vocal characteristics.

Stage 3 — Synthetic output

The model produces:

"I recommend this product."

Stage 4 — Commercial exploitation

The company publishes the audio in advertising.

Each stage may raise different legal questions.

A claimant might therefore plead several independent causes of action.

22. Fraudulent Voice Cloning

The most serious cases involve impersonation.

Example:

A criminal creates an AI voice indistinguishable from a company director and instructs an employee to transfer €500,000.

This may involve:

fraud;

criminal liability;

civil restitution;

negligence;

data protection;

personality rights.

The civil claim might seek recovery from:

the impersonator;

the platform;

an intermediary;

a negligent service provider,

depending upon the facts and national law.

23. Defamatory Voice Cloning

Consider:

AI clone of a politician says: "I accepted a €2 million bribe."

The real politician never said this.

Potential claims may include:

personality rights;

privacy;

defamation;

reputation protection;

injunction;

damages.

The case becomes particularly complicated because the defendant may invoke:

political expression;

satire;

parody;

public interest.

The French 2026 approach to balancing expression and personality rights becomes highly relevant by analogy.

24. Pornographic Voice Cloning

Unauthorized voice cloning in sexually explicit material presents particularly serious personality and dignity concerns.

Potential civil claims may involve:

privacy;

dignity;

personality rights;

image/likeness;

defamation;

emotional harm;

injunctions;

removal orders.

Where the synthetic content falsely suggests participation by the victim, the infringement may be especially serious.

25. Remedies

A claimant can potentially seek several remedies.

1. Injunction

The court can prohibit further use of the cloned voice.

This is often the most important remedy.

2. Removal

The defendant may be required to remove:

videos;

advertisements;

audio files;

social-media posts;

websites.

3. Destruction or disabling of models

This is more technically complicated.

A court could potentially need to consider whether:

a voice model itself unlawfully embodies the claimant's protected identity.

This remains an emerging legal issue.

4. Damages

Damages may compensate:

emotional harm;

reputational harm;

loss of commercial opportunities.

5. Reasonable licence fee

German jurisprudence allows the possibility of calculating economic harm by asking:

What licence fee would reasonably have been payable for authorised use?

This is particularly attractive in celebrity voice cases.

6. Unjust enrichment

Where the defendant has commercially benefited from unauthorised exploitation, the claimant may seek an enrichment-based remedy where national law permits it.

26. Fictitious Licence Fee

This remedy deserves special attention.

Suppose a famous voice actor normally charges:

€25,000

for a commercial voice endorsement.

A company instead creates an AI clone and avoids paying the fee.

The claimant may argue:

"I should receive the reasonable value of the licence that the defendant appropriated."

This prevents the defendant from arguing:

"You cannot prove that you personally lost €25,000."

The economic injury lies in the appropriation of the licensing opportunity.

27. Moral or Non-Pecuniary Damage

Voice cloning can also create non-economic harm.

For example:

A famous actor's AI voice is used to promote a political ideology that the actor publicly rejects.

The financial loss may be difficult to calculate.

But the personality injury may be substantial.

The claimant can therefore potentially seek:

compensation for reputational harm;

emotional distress where national law allows;

injunctive relief;

correction;

publication of judgment.

28. Public Figures Versus Ordinary Individuals

Celebrity cases are not the only relevant cases.

An ordinary person's voice can also be protected.

For example:

An employee's voice is secretly cloned and used in a fraudulent customer-service chatbot.

The person may not have significant commercial publicity value.

Nevertheless, the claim may be strong because of:

privacy;

autonomy;

identity;

deception.

Celebrity status therefore affects economic value, but does not necessarily determine whether a voice is legally protected.

29. Publicly Available Voice Does Not Mean Free to Clone

A common defence is:

"The person's voice was already available on YouTube."

That is not necessarily sufficient.

There is a fundamental distinction between:

public availability

and

permission for unrestricted AI exploitation.

A person may voluntarily upload a speech while retaining rights concerning:

commercial exploitation;

cloning;

synthetic reproduction;

false attribution.

The French 2026 case is particularly relevant because public availability of speech did not eliminate the need to balance the speaker's voice rights against expressive freedom.

30. Voice Imitation Versus Exact Recording

There are two different legal situations.

Exact recording

The defendant takes:

"Hello, this is John."

from a genuine recording.

Synthetic imitation

The defendant creates:

"I recommend this product."

using AI that sounds like John.

The second case may actually be more difficult technologically but does not necessarily escape personality-right protection.

The Austrian and German cases demonstrate that imitation itself can be legally significant.

31. Causation

For damages, the claimant may need to establish:

unauthorised voice use → legal injury → financial or non-financial damage.

Examples:

Lost endorsement

Celebrity normally charges €100,000.

Reputational injury

AI voice associates person with an offensive statement.

Market confusion

Consumers believe the person endorsed a product.

Lost professional opportunities

A voice actor's clients stop hiring them because the AI clone is being used instead.

32. Evidence

Voice-cloning litigation will be highly evidence-intensive.

Important evidence includes:

Audio analysis

Experts can compare:

timbre;

pitch;

formants;

cadence;

pronunciation;

spectral characteristics.

Audience evidence

Statements from viewers may establish:

"I believed the real actor was speaking."

Commercial records

advertising revenue;

licensing fees;

endorsement rates;

YouTube monetisation;

subscription income.

Technical records

model training records;

prompts;

generation logs;

source recordings;

model documentation.

Contractual records

talent agreements;

recording agreements;

AI clauses;

licensing contracts.

33. Defences

A defendant may raise several arguments.

A. Consent

"The claimant authorised the use."

B. Lack of recognisability

"The voice does not identify the claimant."

C. Artistic freedom

"This was satire or artistic expression."

D. Public interest

"The use contributed to a public debate."

E. Parody

"The purpose was humorous imitation rather than commercial exploitation."

F. No economic harm

"The claimant cannot prove financial loss."

This defence is weaker where the law permits a licence-fee or personality-right remedy.

34. Why Commercial Use Is Particularly Risky

Consider two examples.

Example 1 — Political parody

AI imitates a politician's voice in an obvious satirical video.

There may be strong freedom-of-expression arguments.

Example 2 — Advertisement

AI imitates the same politician's voice saying, "I personally recommend this investment."

There is:

commercial exploitation;

likely deception;

implied endorsement;

potential reputational injury.

The second case is much more legally problematic.

35. European Comparative Position

36. Important Distinction: AI Act Does Not Replace Civil Law

The European AI Act primarily addresses:

transparency;

AI-system obligations;

synthetic-content marking;

deepfakes;

regulatory compliance.

It does not create a comprehensive EU-wide property right in a person's voice.

Therefore:

AI Act + national personality law + GDPR + copyright + contract law

may all operate simultaneously.

37. Six Most Important Cases at a Glance

38. Six Core Legal Principles

Principle 1 — Voice can be a personality right

A person's voice can be protected independently of copyright.

Principle 2 — Imitation can be actionable

Protection does not necessarily depend upon copying an original audio file.

Principle 3 — AI does not eliminate liability

Replacing a human impersonator with an algorithm does not automatically eliminate personality-right protection.

Principle 4 — Commercial exploitation is especially problematic

Unauthorised advertising and commercial endorsement create a strong basis for civil claims.

Principle 5 — Identification matters

The more recognisable the cloned voice, the stronger the personality-right argument generally becomes.

Principle 6 — Expression and personality rights must be balanced

Satire, parody, journalism and artistic expression can justify some uses, but commercial deception is much harder to defend.

39. Hypothetical Example

Assume a famous European actor has a distinctive voice.

An AI company obtains 30 hours of publicly available interviews.

It creates a synthetic voice model.

A cryptocurrency company then pays €50,000 to use that model in advertisements saying:

"I personally recommend this cryptocurrency."

The actor never authorised the use.

The advertisement reaches five million people.

The token subsequently collapses.

Potential civil claims

The actor may potentially claim:

infringement of personality rights;

unauthorised commercial exploitation of voice;

false endorsement;

reputational harm;

injunction;

removal;

damages;

reasonable licence fee;

unjust enrichment where available;

data-protection remedies if personal/biometric data was unlawfully processed.

The consumers who relied on the advertisement may separately have claims based on:

consumer law;

fraud;

misrepresentation;

financial regulation.

Thus, one cloned voice can generate several categories of litigation simultaneously.

40. A Model Civil-Law Test

For European voice-cloning disputes, a useful analytical framework is:

Question 1

Is the voice recognisable?

↓

Question 2

Was the use authorised?

↓

Question 3

Was an original recording copied, or was a synthetic imitation created?

↓

Question 4

Does national law recognise a personality interest in the voice?

↓

Question 5

Was the use commercial, artistic, journalistic, political or private?

↓

Question 6

Did the use create confusion or false endorsement?

↓

Question 7

Are privacy or GDPR rights also implicated?

↓

Question 8

Does copyright or a related right protect the underlying recording/performance?

↓

Question 9

Does freedom of expression or artistic freedom justify the use?

↓

Question 10

What remedy is appropriate?

41. Conclusion

Unauthorized voice cloning is becoming a distinct European civil-law problem, although European law still largely approaches it through existing personality, privacy, contract, copyright, data-protection and unfair-commercial-practice doctrines rather than through a single harmonised "voice ownership" right.

The strongest modern authorities are the Berlin Regional Court's 2025 AI voice decision and the French Court of Cassation's 24 June 2026 decision, while the Austrian 6 Ob 287/02b and German OLG Hamburg 3 W 45/89 cases demonstrate that protection against recognisable voice imitation existed well before generative AI.

The developing legal position can be summarised as follows:

A person does not necessarily own every sound resembling their voice, but a sufficiently recognisable reproduction or imitation of their distinctive voice can constitute an unlawful appropriation of a protected personality attribute, particularly where it is used commercially, creates false endorsement, damages reputation, or exploits the person's economic identity without consent.

The 2026 French decision is particularly significant because it expressly treats the voice alongside the image as a protected personality attribute while requiring courts to balance that right against freedom of expression. The 2025 German AI-voice decision is equally significant because it applies the personality-right concept directly to an AI-generated voice imitation.

Finally, as of 2 August 2026, EU AI Act transparency rules concerning relevant synthetic audio and deepfakes are applicable. These rules strengthen transparency against deceptive synthetic media, but labelling an AI-generated voice does not by itself grant permission to exploit another person's identity. Civil personality rights therefore remain independently important. (EUR-Lex)