Voice Cloning Laws in 2026: US, EU, UK and Beyond

Voice cloning laws depend on where you are. Cloning with the speaker's consent is the lawful path, while unauthorised cloning can breach several laws. Tennessee and California have voice statutes in force, the US NO FAKES Act is still a bill, and EU transparency rules have applied since 2 August 2026. This is general information, not legal advice.

Is voice cloning legal?

The answer turns on consent, the purpose of the clone and the jurisdiction. Most rules target unauthorised replicas of a real person's voice, misleading uses that could fool listeners, and robocalls. Fewer rules address a person who clones their own voice for personal use, though platforms such as Apple limit their own features to personal, non-commercial use. Read the statutes that apply to you, because the sources below are summaries and rules change.

Voice cloning laws at a glance

| Jurisdiction | Instrument | Status (17 October 2026) | Key date | |---|---|---|---| | US federal | NO FAKES Act (S.4591 / H.R.8915) | Bill, stalled | Introduced 20 May 2026; committee markup June 2026 | | US federal | FCC ruling on AI voices in robocalls | In force | 8 February 2024 | | Tennessee | ELVIS Act | In force | Effective 1 July 2024 | | California | AB 2602 and AB 1836 | In force | Effective 1 January 2025 | | Illinois | HB 4875 (Public Act 103-0836) | In force | Effective 1 January 2025 | | Washington | SSB 5886 | In force | Effective 11 June 2026 | | Montana | HB 513 | In force | Effective 1 January 2026 | | EU | AI Act Article 50 | Applies | From 2 August 2026 | | Denmark | Copyright deepfake bill | Proposed, not law | Introduced 8 October 2026 | | UK | No dedicated personality right | Not law; consultation promised | March 2026 pledge | | China | AI content labeling measures | In force | Effective 1 September 2025 | | India | Court orders (Bombay High Court) | Case law | 2024 |

United States: the federal picture

The NO FAKES Act would give individuals a federal, licensable property right in their voice and visual likeness. The 2026 version was reintroduced on 20 May 2026 by a bipartisan group, with Senate sponsors including Sen. Blackburn, Sen. Coons, Sen. Tillis and Sen. Klobuchar. The Manatt client alert lists the new features:

A DMCA-style counter-notification process, with content restored no sooner than 14 days after a counter-notice unless the complainant sues.
A penalty for knowingly false counter-notifications of the greater of $25,000 per notice or actual damages plus fees.
An exclusion for nonprofit libraries, archives and accredited nonprofit educational institutions.

The Senate Judiciary Committee took the bill up at a markup in June 2026, and trade reports say it advanced; check Congress.gov for the official record. House Judiciary has not acted on the companion bill. On 30 September 2026, Sen. Ted Cruz objected to a unanimous-consent request to pass the bill, and it is now stalled. As of October 2026, the NO FAKES Act is still a bill, not law.

The FCC ruling of 8 February 2024 is in force. It holds that AI-generated or simulated voices count as an "artificial or prerecorded voice" under the TCPA, so callers need prior express consent unless an exemption applies. Law firms read the ruling differently. Some say it applies existing consent rules rather than banning the technology.

United States: state laws

States have moved faster than Congress. These are the laws the sources reviewed describe:

Tennessee ELVIS Act: signed on 21 March 2024 and effective 1 July 2024. It protects a voice whether the sound is the actual voice or a simulation. Liability can reach anyone who distributes a tool whose primary purpose is producing a specific person's voice or likeness, if they know the use is unauthorised. Comment, criticism, scholarship, satire and parody have exemptions. See the Manatt summary of the ELVIS Act.
California AB 2602 and AB 1836: effective 1 January 2025. AB 2602 makes some contract clauses allowing AI replicas unenforceable, and AB 1836 creates liability for unauthorised digital replicas of deceased personalities, with a floor of $10,000 or actual damages.
Illinois HB 4875: effective 1 January 2025. It makes unauthorised AI digital replicas actionable, with facilitators liable only with actual knowledge.
Washington SSB 5886: effective 11 June 2026. It adds "forged digital likeness" to protected attributes and raises the civil penalty to $3,000. It is a civil measure, not a criminal one.
Montana HB 513: effective 1 January 2026. It creates a property right in name, voice and visual likeness that lasts 20 years after death.
Pennsylvania Act 35 of 2025: effective 5 September 2025. It creates a criminal "digital forgery" offence.

New York's synthetic performer disclosure law covers advertisements, but it does not apply to audio advertisements, so it is not a voice-cloning rule.

European Union: AI Act Article 50

The EU AI Act's Article 50 took effect on 2 August 2026. Two duties matter most for voice:

Providers must ensure synthetic audio, image, video and text outputs are "marked in a machine-readable format and detectable as artificially generated or manipulated," as far as technically feasible.
Deployers of AI that generates deepfakes must disclose that the content is artificially generated or manipulated. The duty is lighter for evidently artistic, creative, satirical or fictional works.

Under the Digital Omnibus on AI, Regulation (EU) 2026/1744, generative systems placed on the market before 2 August 2026 have until 2 December 2026 to implement the machine-readable marking. Systems placed on the market on or after 2 August 2026 get no grace period. These dates come from law-firm summaries, so check the official text. Fines for transparency breaches can reach EUR 15 million or 3% of worldwide annual turnover, whichever is higher, with the lower figure applying to SMEs and start-ups.

Denmark's proposed law

Denmark's culture minister, Zenia Stampe, introduced a bill on 8 October 2026. It would amend the copyright law to prohibit sharing realistic deepfakes of a person's face or voice without consent, covering ordinary people as well as performers. Satire, parody, caricature and criticism would remain allowed, except where the content is dangerous misinformation. The minister hopes the law takes effect at the start of 2027. The Euronews report describes it as a proposal, not law.

United Kingdom

The UK has no codified personality right. Protection is a patchwork of passing off, copyright in a specific recording, performers' rights, data protection, privacy and defamation. The government pledged in March 2026 to consult on options, including a specific personality right. As of the Pinsent Masons analysis of 15 September 2026, the consultation was delayed and unconfirmed.

China

China's Measures for Labeling of AI-Generated Synthetic Content were issued on 14 March 2025 by four national agencies and took effect on 1 September 2025, with accompanying national standards. They require explicit labels, including for audio, and implicit labels embedded in file metadata. The standard gives examples of audio labels, such as an "AI" notice spoken through voice and rhythm.

India

Indian courts have acted through case law rather than statute. In Arijit Singh v. Codible Ventures LLP (2024), the Bombay High Court held that offering AI tools that convert any voice into a celebrity's voice without permission violates the celebrity's personality rights. The case is reported in a WIPO Magazine article.

How to stay on the right side of the law

Whatever jurisdiction applies, these practices reduce risk:

Get written consent from the speaker that names the purpose of the clone.
Keep the clone within the scope you agreed, and do not reuse it elsewhere.
Label AI-generated audio, especially when listeners could mistake it for a live person.
Never impersonate a real person to deceive a listener or customer.
Keep records of consent, originals and where the audio was published.

How OndaLabs fits

OndaLabs' custom voices are for recordings you have permission to use, and that principle matches the consent steps above. OndaLabs' custom voices cost 1,500 credits to clone or design, and slots run from 3 on Creator to 30 on Business. OndaLabs does not make legal claims for you, and the product facts here do not say what labelling or verification it applies. For the consent process, read how to clone your voice. For the custom voices page, see OndaLabs custom voices, and compare plans on the pricing page.

Frequently asked questions

Is it legal to clone someone's voice?

It depends on consent, purpose and location. Cloning with the speaker's written consent is the lawful path, while unauthorised cloning can breach state laws such as Tennessee's ELVIS Act or California's AB 2602 and AB 1836. This is general information, not legal advice.

Has the NO FAKES Act passed?

No. As of 17 October 2026 the NO FAKES Act (S.4591 and H.R.8915) is still a bill. It went through a Senate Judiciary Committee markup in June 2026, and Sen. Ted Cruz objected to a unanimous-consent request on 30 September 2026.

Does the EU AI Act require labelling of AI-generated voices?

Yes, in the form described in Article 50. Providers must mark synthetic audio and other outputs in a machine-readable way, and deployers of deepfake-generating systems must disclose that content is artificially generated. Generative systems already on the market have until 2 December 2026 to add the marking, according to law-firm summaries.

Are AI voice robocalls illegal?

Not automatically, but they need prior express consent under the TCPA unless an exemption applies. The FCC's February 2024 ruling treats AI-generated voices as artificial or prerecorded voices, so the existing consent rules apply.

Is voice cloning legal in the UK?

The UK has no codified personality right. Protection comes from several separate areas of law, and a dedicated right has been promised for consultation. Check current guidance before you clone or publish a voice in the UK.

Do I need consent to clone a voice?

Most vendors require it. ElevenLabs asks for confirmation of consent, Microsoft requires written permission from the voice talent, and OpenAI requires a recorded consent statement from the same person whose voice is cloned.