Who Owns Your Voice? Voice Rights Law, Country by Country

Nowhere in the world can you copyright your own voice. That is the starting point most people get wrong. What protects a voice actor against an unauthorised clone is personality law, publicity law or unfair competition law — and the answer changes completely depending on which country the recording is published in.

One first-instance German judgment has now put a number on it: €2,000 per video for an AI imitation of a recognisable dubbing voice. Denmark is trying to move voice protection into copyright law altogether. The United States has fifty different answers and one pending federal bill. This page collects what each jurisdiction actually says, with the primary source behind every claim.

The short answer: protected, but not as property

Copyright protects works — a recorded performance, a script, a mix. It does not protect the timbre of a human voice, because a voice is not an authored work. So when someone trains a model on your recordings and generates new speech in your voice, copyright in the source recordings is only half the question, and often the weaker half.

The stronger claim usually sits somewhere else: a personality right in continental Europe, a right of publicity in the United States, passing off in the United Kingdom, and data protection law almost everywhere. These are not interchangeable. A personality right protects you because you are a person; a publicity right protects the commercial value of your identity; passing off protects against a false impression of endorsement. Which one applies determines what you can actually demand — a takedown, a licence fee, or nothing at all.

Two practical consequences follow. First, protection is territorial: a clone that is unlawful in Berlin may be lawful in London. Second, in most jurisdictions your contract matters more than the statute, because the statute usually asks whether you consented.

Germany: the right to your own voice, with a price attached

German law has no equivalent of the image-rights provisions in sections 22 and following of the Kunsturhebergesetz for voice. The voice is instead protected as an aspect of the general personality right, derived from the constitution and enforced through section 823(1) of the German Civil Code, with courts applying the image-rights rules by analogy.

On 20 August 2025 the Landgericht Berlin II decided the case that voice actors had been waiting for. A YouTuber had used an AI-generated imitation of Manfred Lehmann — the German dubbing voice of Bruce Willis and Gérard Depardieu — to narrate two videos, without asking. The court held this to be an unlawful interference with Lehmann’s personality right and awarded a notional licence fee of €2,000 per video, €4,000 in total (case no. 2 O 202/24; reported by Legal Tribune Online, with the full text published by a Berlin law firm and the case indexed at dejure.org).

Three findings in that judgment matter beyond the individual case. The court said it makes no difference whether a recognisable voice is imitated by a human impressionist or generated by a model. It rejected the satire defence because the videos pointed to the defendant’s online shop, which made the use commercial. And it treated the absence of any AI labelling as part of the problem, because viewers could reasonably assume Lehmann had agreed.

The figure was estimated under section 287(2) of the German Code of Civil Procedure, taking into account how well known the voice is and the channel’s reach of roughly 190,000 subscribers. Two cautions before anyone treats €2,000 as a market rate: this is a first-instance regional court, not the Federal Court of Justice, so it binds nobody, and the amount was tied to the fame of one of the most recognisable voices in German dubbing. The underlying principle is older — the Federal Court of Justice accepted in its 1999 Marlene Dietrich line of cases that personality features carry economic value that survives unauthorised use.

The court added a data-protection layer: if a cloned voice is treated as personal data, the balancing test under Article 6(1)(f) GDPR did not favour the YouTuber, and the journalistic and artistic exemption in Article 85 GDPR did not apply either.

Denmark: the attempt to put voice inside copyright

Denmark is the outlier. Rather than strengthening personality law, the Danish government has drafted amendments to its Copyright Act that would give every natural person control over realistic digitally generated imitations of their personal characteristics, voice included, with a separate layer of protection for performers.

The European Parliamentary Research Service briefing of January 2026 sets out the mechanics: consent is required to make such content available, protection lasts 50 years after death, and enforcement runs through the notice-and-action procedure in Article 16 of the Digital Services Act. Caricature, satire, parody and pastiche are in principle excluded.

The same briefing is unusually candid about the weaknesses. The Danish authorities themselves state in the explanatory notes that they do not intend to create a new copyright for citizens — the amendments sit in the Copyright Act but function as codified personality rights. And the effect is geographically limited: very large platforms would geo-block the content for Denmark, leaving it reachable everywhere else unless other member states follow.

On timing, be careful. The draft was notified to the European Commission and open for comments until 3 February 2026, and the EPRS briefing recorded an expected entry into force in July 2026. Earlier reporting had pointed to 31 March 2026, before a snap general election intervened. Anyone relying on the Danish rules should check the current status of the adopted text rather than the announcements.

The direction of travel is clearer than the Danish timetable. A declaration agreed under the Danish Council presidency, signed by every member state except Hungary plus five other European countries, includes the principle that citizens must be protected against digital replicas of their personal characteristics without consent.

The United States: fifty answers and one pending bill

There is no federal right of publicity. Voice protection in the US is state law, and it varies in scope, duration and remedies.

Tennessee moved first. Governor Bill Lee signed the ELVIS Act on 21 March 2024, effective 1 July 2024, extending the state’s publicity right expressly to voice and to tools whose primary purpose is unauthorised imitation. It is codified at Tenn. Code Ann. section 47-25-1101 and following.

California took a contractual route, which is the more useful model for working voice actors. Assembly Bill 2602 added section 927 to the Labor Code, effective 1 January 2025. A contract clause permitting a digital replica of someone’s voice or likeness to replace work they would otherwise have performed is unenforceable unless the intended uses are described with reasonable specificity and the performer was represented by counsel or covered by a collective agreement addressing digital replicas. Its companion, AB 1836, extends consent requirements to replicas of deceased performers.

At federal level the NO FAKES Act would create a national digital replication right covering voice and visual likeness, with platform liability for knowingly hosting unauthorised replicas and a notice-and-takedown structure. The Senate Judiciary Committee advanced S. 4591 unanimously on 18 June 2026 and the bill was reported with an amendment; the committee text is on congress.gov. It is not law. It still needs a Senate floor vote, House passage and signature, and the reported figure of up to $750,000 per work is a proposal, not a statutory penalty.

Around those three, the picture is a patchwork rather than a system. States including California, New York and Illinois apply pre-existing publicity, privacy or biometric statutes to synthetic voices, while a separate wave of state laws targets sexual and election deepfakes specifically and does nothing for a commercial voice clone. Two questions decide whether a US claim is worth bringing: whether the state protects voice as such rather than only name and image, and whether the right survives death, which matters enormously for archive and estate work.

The United Kingdom: no personality right, and a consultation instead

The UK recognises no image right, character right or personality right. The nearest tool is passing off, which requires goodwill, a misrepresentation and damage — the route Eddie Irvine used successfully against Talksport in 2002 on a false endorsement theory. For a voice clone that carries no implication of endorsement, that route often fails.

Reform is being discussed rather than delivered. The government consultation on copyright and AI closed on 25 February 2025 with more than 11,500 responses, and the government published its report on 18 March 2026. The report acknowledged that existing law does not protect identity against a synthetic replica generated from a performer’s back catalogue, and set out options ranging from a narrow digital replica right to a broader personality right, with a dedicated consultation promised for summer 2026. The House of Lords Communications and Digital Committee published its own recommendations on 6 March 2026 (report, HL Paper 267).

Until something is enacted, a British voice actor’s strongest protection is the contract they sign, not the statute book.

The EU layer: disclosure is not ownership

It is worth separating two things that are routinely conflated. The EU AI Act creates transparency duties, not rights in a voice. Article 50 has applied since 2 August 2026 and requires that synthetic audio be machine-readably marked and, for deepfakes as defined in Article 3(60), disclosed. It tells you what must be labelled; it does not tell you who may license a voice. We cover the operational detail in our guide to Article 50 and synthetic voice disclosure.

The rights question in the EU still runs through national personality law and through the GDPR, where a voice recording is personal data and, depending on the processing, potentially biometric data. That combination is why the Berlin court could reach a data-protection conclusion alongside a personality-rights one.

One more distinction is worth holding onto, because it is where most disputes actually start. Training a model on existing recordings and publishing synthetic speech generated from that model are two separate acts, and they can fall under different rules. Training may implicate the rights in the recordings and, in the EU, the text-and-data-mining framework; publication implicates the personality or publicity right of the person the output sounds like. A licence covering one does not cover the other, which is exactly the ambiguity that most AI clauses in voice contracts exploit.

What contracts do that statutes cannot

Every framework above turns on consent. That makes the contract the operative document, and collective agreements have moved faster than legislatures.

SAG-AFTRA’s 2025 Interactive Media Agreement, ratified in July 2025 after a strike of nearly a year, requires consent and disclosure for digital replica use and lets performers suspend consent for the generation of new material during a strike. The union maintains a public summary of its AI provisions across contracts.

In Germany the same question is being fought outside the courts. A large part of the dubbing industry has refused Netflix work since February 2026 over a clause covering the use of recordings for AI training — we track the state of that dispute in a separate Netflix and German dubbing page.

The practical lesson from both: statutes decide what happens when there is no agreement, which is the worst place to find yourself. Rates and usage terms are the same conversation — see our comparison of union rates in the US, UK and Germany.

If you record or commission voice, do these five things

• Name the jurisdiction. Governing law is not boilerplate here. It decides whether you have a personality right, a publicity right, or passing off.
• Separate the two grants. A licence to use a recording is not a licence to train on it. Write synthetic use as its own clause, with its own fee and its own term.
• Describe the uses specifically. California now makes vague replica clauses unenforceable. Specificity protects both sides.
• Keep the consent revocable in defined cases. The SAG-AFTRA suspension right is a template worth borrowing.
• Label synthetic audio. Since 2 August 2026 marking is a legal duty in the EU for material in scope, and the Berlin court treated missing labelling as an aggravating factor.

None of this is legal advice, and the position in several of these jurisdictions will have moved by the time you need it. What has not changed is the commercial logic: a recognisable voice has measurable value, and courts are now willing to put a figure on unauthorised use.

If you would rather sidestep the ambiguity, the cleanest route is a booking where the usage terms are written down before recording starts. That is how projects run on Voicfy — native-language talent, agreed usage, and a human voice on the file.