British performers are pressing the government to give individuals stronger legal control over artificial-intelligence systems that reproduce their voices, adding pressure to a problem ministers have already acknowledged is not fully covered by existing copyright and performer protections. Around 80 performers and supporters have backed the Save Our Voices Now campaign, which is seeking legislation intended to prevent voices being replicated and used commercially or criminally without consent.

The campaign includes prominent actors and voice performers and has taken its demands to government alongside a Parliamentary petition seeking stronger legal protection around individual voices. The intervention places performers within a wider debate over whether Britain needs rights designed specifically for realistic digital replicas rather than relying on a patchwork of laws created for recordings, personal data, fraud, endorsement, and other forms of harm.

That debate acquired a firmer policy basis in March, when the government concluded that existing legislation can leave people without meaningful control over synthetic copies of their voice or appearance. Its report on copyright and artificial intelligence said AI had made convincing replicas cheaper and easier to create, while Britain’s combination of copyright, performers’ rights, data protection, criminal law, passing off, and other civil remedies did not cover every instance of unauthorised imitation.

The gap exists partly because copyright generally protects a recorded creative work rather than the characteristics of the person whose voice appears in it. An AI-generated imitation can sound recognisably like an actor without necessarily reproducing a substantial part of one protected sound recording, while the individual being imitated may not own the copyright in material from which the synthetic likeness was developed.

A voice is not straightforwardly copyright

The government’s analysis makes that distinction explicit. Britain has no general personality right and no dedicated intellectual-property right covering digital replicas. Existing law can apply in particular circumstances, although a person discovering an unauthorised synthetic voice may need a different legal route depending on whether the case involves copied recordings, personal data, fraud, defamation, or false endorsement.

Performers have additional rights relating to recordings and broadcasts of their performances, but those rights do not amount to general ownership of the sound of an individual’s voice. Contracts complicate the position further because economic rights in performances can be licensed or assigned, potentially reducing the performer’s direct control over recordings that later contribute to AI use.

Generative systems stretch those distinctions because they can learn characteristics from recorded material and produce speech the person never performed. The result may preserve enough vocal identity to be commercially valuable without directly reproducing an existing performance, leaving law designed around copying fixed works with an awkward problem.

The commercial consequences extend beyond well-known actors. Narrators, translators, audiobook readers, advertising performers, game actors, and other workers can derive income from a recognisable voice that synthetic systems are becoming increasingly capable of approximating. Consent therefore becomes central when companies want to create digital doubles, localise productions, reproduce dialogue, or generate additional material from existing performances.

Government has already accepted a protection gap

Rather than concluding that copyright should simply be stretched to make every person’s voice an owned work, ministers said they would examine broader protections. One possibility is a digital-replica or personality right giving individuals more control over commercial reproduction of their voice, face, or likeness, although consultation has not produced a settled view on the scope of such a right or who should carry liability when it is breached.

Those design questions become difficult quickly. Liability could sit with an AI developer, the organisation commissioning an imitation, the person generating it, or the platform distributing it, while legitimate applications include dubbing, accessibility technology, authorised digital doubles, entertainment production, and synthetic speech for people who have lost their natural voice.

Effective protection also has to be usable. A performer discovering an unauthorised advert or synthetic narration gains little from a theoretical right if identifying the responsible party and securing removal requires litigation costing more than the lost work.

Other jurisdictions are already experimenting with different approaches, including protections aimed specifically at digital likenesses or commercial personality rights. Britain could strengthen performers’ rights, establish a dedicated replica right, create broader personality protection, or combine elements of several models.

Fraud adds a separate problem that copyright was never intended to resolve. Synthetic voices can impersonate relatives, executives, employees, or public figures to obtain money or confidential information, where the commercial value of the voice lies in persuading somebody that a real person is speaking rather than exploiting a creative performance.

Businesses deploying voice-generation technology therefore have reasons to tighten consent and provenance controls before Parliament chooses a legislative model. Recording contracts, employment agreements, model-training permissions, advertising terms, and synthetic-media policies can determine whether an organisation can show that a reproduced voice was authorised and that the individual understood the uses being licensed.

The government’s own report means the policy question has moved beyond whether voice cloning exposes gaps in current law. The unresolved issue is how protection should be constructed, how broadly it should extend beyond professional performers, and whether individuals can obtain practical control without blocking legitimate synthetic-speech applications built from the same underlying technology.