Short answer: yes. There is no UK law that bans creating an AI copy of a voice. What matters is whose voice it is, whether they agreed, what you use it for, and whether anyone is misled. A business that clones a willing person's voice under a clear written agreement, and is open with listeners that the voice is synthetic, is on solid ground.
This guide explains the rules that apply in plain terms. It is a general overview as of October 2026, not legal advice. For anything unusual, such as a celebrity voice, a deceased person, or an advertising campaign, take advice from a solicitor.
What voice cloning is
Voice cloning is the process of training a synthetic voice on recordings of a real person, so that software can then read out any text in a voice that sounds like them. Businesses use it for AI phone agents, training videos, product narration, accessibility, and translating content into other languages in the same voice.
The laws that apply
Several separate areas of UK law touch voice cloning. None of them is about voice cloning specifically, which is why the subject feels unclear.
- UK GDPR and the Data Protection Act 2018. A recording of an identifiable person's voice is personal data. You need a lawful basis, a stated purpose, a retention period, and a way to respond if the person asks for their data. Where voice data is used to identify someone, it can count as biometric data, which is a special category with stricter conditions.
- Performers' rights under the Copyright, Designs and Patents Act 1988. Someone who performs, for example a voice artist recording a script, has rights over recordings of that performance. Copying those recordings, including to train a voice, needs their consent.
- Copyright in the recordings themselves. Whoever owns a sound recording controls copying it. Training on audio you do not own, such as clips from a podcast or a film, can infringe that copyright even if the speaker is happy.
- Passing off and false endorsement. The UK has no general personality right, but using a recognisable voice to suggest someone endorses your product can be passing off. This is the main risk with famous voices.
- Fraud and harassment law. Using a cloned voice to deceive someone for gain is fraud, whatever the technology. Using it to harass or impersonate someone is also unlawful.
- Advertising rules. The UK advertising codes prohibit misleading ads and implying an endorsement that does not exist, which covers synthetic voices in adverts.
- Calling rules. If the voice is used for outbound automated calls, the Privacy and Electronic Communications Regulations apply, and automated marketing calls need prior consent.
If you sell into the EU, also note the EU AI Act. From August 2026 it requires that audio content generated or manipulated by AI to resemble real people is disclosed as such, with limited exceptions.
What a voice agreement should cover
Almost every legal risk above is handled by one document: a written agreement with the person whose voice is being cloned, signed before anything is recorded. It should cover at least:
- 01Who the voice belongs to and who the business is.
- 02The specific uses allowed, for example the company phone line and training videos, and the uses that are not.
- 03The term: how long the voice may be used, and what happens when the person leaves the business.
- 04Languages and channels the voice may appear in.
- 05Payment, if any, and whether it is a one-off fee or ongoing.
- 06How the person can withdraw consent, and a commitment to delete the voice model when they do.
- 07Where the recordings and the voice model are stored, who can access them, and how long the raw recordings are kept.
For staff, keep the voice agreement separate from the employment contract. Consent buried in a contract of employment is weak, because the person may not feel free to refuse.
Uses that cross the line
- Cloning anyone without their documented consent, including from public clips of them speaking.
- Imitating a celebrity, politician, or other public figure to suggest they said or endorse something.
- Phone calls designed to make the listener believe they are speaking to a real person who is not there.
- Keeping and using the voice of someone who has left the business or withdrawn consent.
Good practice that protects you
- Tell listeners the voice is AI generated, particularly on phone calls. A short line in the greeting is enough.
- Store the voice model on an account your business controls, with access limited to the systems that need it.
- Log where and when the voice is used.
- Keep the signed agreement with the voice files, so you can show consent if asked.
- Use a stock synthetic voice when you do not specifically need a real person's voice. It needs no consent paperwork at all.
We build custom AI voices on this basis: agreement first, recording second, and an off switch that actually deletes the voice. If you are weighing up a cloned voice for a phone agent or content, we can tell you whether a stock voice would do the job instead.