AI voice cloning uses machine learning to imitate a real person’s speech. SoaplandTV reports that Derry Girls stars are backing a campaign on AI and voices, and that its petition has passed 10,000 signatures.
AI voice cloning defined
AI voice cloning is the use of machine-learning software to produce new speech that sounds like a particular person. Unlike an edited recording, the output can contain words the person never said. The result is usually called a synthetic voice or voice clone. When it is used to deceive, it is often called an audio deepfake.
The subject has reached a wider audience through a report from SoaplandTV, which says that stars of Derry Girls, the Channel 4 comedy set in Northern Ireland, are supporting a campaign about AI and voices. It also says the campaign’s petition has passed 10,000 signatures. The summary available for this article does not give the organisers, the wording of the petition, the cast members involved or where it is hosted, so those details are treated here as unknown.
The wider issue can be explained without them. It has two sides: a question of rights, about who controls a voice once software can reproduce it, and a question of security, about what happens when people or systems trust a voice that is not genuine.
Origins of the technology and the dispute
Computers have been producing speech for decades. Early systems generated sound according to hand-written rules and were easy to recognise as artificial. Later systems stitched together small fragments of recorded speech, which sounded more natural but needed extensive recording sessions with a single speaker.
Neural networks changed the economics. Models trained on large collections of speech learned the general patterns of human voices, and could then be adapted to a specific speaker using far less material. Tools that once belonged to research laboratories became commercial services, some of them available to anyone with a web browser.
That shift created the current dispute. A performer’s voice is part of their professional livelihood, and many performers have a large body of recorded work in circulation. Performers’ organisations, including the UK union Equity, have raised concerns about consent, payment and the use of recordings to train AI systems. Fraud warnings followed a parallel track, as criminals found that a familiar voice could make a scam call more persuasive.
How voice cloning works today
Most voice cloning follows a similar sequence. Audio of the target speaker is collected and processed so that a model can capture characteristics such as pitch, rhythm and accent. The user then supplies text, which the system turns into speech in that voice, or supplies a recording of their own speech, which the system converts. Quality varies with the source audio, the software and the amount of human editing afterwards.
Legitimate uses exist. They include dubbing, audio production, accessibility tools and voice banking for people who expect to lose the ability to speak. Many commercial providers state that they require consent from the speaker, although the strength of those checks differs between services.
The legal position in the UK is piecemeal. There is no single statute giving a person general ownership of the sound of their voice. Performers’ rights under the Copyright, Designs and Patents Act protect recorded performances, while the common-law claim of passing off may help where a false endorsement is implied. Data protection law can also apply, since voice recordings can be personal data. Each route has limits, which is why campaigners and unions have argued for clearer protection. Other jurisdictions have legislated more directly: the US state of Tennessee has passed a law covering voice likeness, and the EU’s AI Act includes transparency duties for deepfakes.
On the security side, the main risks are impersonation and social engineering. A cloned voice can support a fraudulent call claiming to come from a relative, a colleague or a senior executive. It can also put pressure on organisations that use voice recognition to authenticate customers, which is one reason security practitioners advise against relying on voice as the only check.
Common misconceptions
A voice is not automatically protected by copyright. Copyright and related rights attach to works and recorded performances. A new clip generated to sound like someone is a harder case, because it may not copy any specific recording.
Consent to a recording is not necessarily consent to cloning. Whether an old contract permits AI use depends on its wording, and many such contracts were written before the technology was in common use. Performers and their representatives are increasingly negotiating this point explicitly.
The issue does not only affect famous people. Celebrity campaigns attract attention, but voice actors, presenters and other working performers can lose income, and ordinary people can be impersonated using audio posted online.
Synthetic speech is not always obvious. The idea that clones sound robotic is out of date for many tools. Detection software exists, but it should not be treated as conclusive, and a poor phone line can hide the flaws a listener might otherwise notice.
A petition does not change the law. Its signatures show support, not legal effect. The UK Parliament’s official petitions site sets signature thresholds for a government response and for consideration of a debate, but it is not known whether this petition is hosted there, so no such consequence should be assumed.
Not all voice synthesis is abuse. The central dispute concerns consent, disclosure, payment and deception, rather than the existence of the technology itself.
Where to look next
Readers following the campaign should start with the petition text itself, which will state what is being asked and of whom, and then check for statements from the organisers. Equity publishes material on AI and performers’ rights. The Intellectual Property Office is the best starting point for government policy on copyright and AI, and the Information Commissioner’s Office explains how data protection law treats personal and biometric data.
For the security side, the National Cyber Security Centre publishes practical guidance on phishing and social engineering. The most useful household and workplace habit is to verify any urgent request made by voice through a second, known channel, such as calling back on a saved number, especially where money or login details are involved.
Frequently asked questions
What is AI voice cloning?
AI voice cloning is the use of machine-learning software to generate new speech that sounds like a specific person. The system is built or adapted using recordings of that speaker, then produces words from typed text or converts another person’s speech. The output can include things the speaker never said.
What is the Derry Girls AI voice petition about?
SoaplandTV reports that stars of Derry Girls are backing a campaign on AI and voices, and that its petition has passed 10,000 signatures. The summary available for this article does not set out the exact demands, the organisers or where the petition is hosted, so readers should check the petition text directly.
Is it illegal to clone someone’s voice in the UK?
There is no single UK law that bans voice cloning as such. Legality depends on the circumstances. Performers’ rights, passing off, data protection law and fraud law may each apply, depending on how the voice was obtained and used. Anyone facing a specific dispute should take legal advice from a qualified professional.
How can I tell if a phone call uses a cloned voice?
Listening alone is not a reliable test, because many synthetic voices sound convincing and phone lines hide flaws. Treat urgency, secrecy and requests for money or login details as warning signs. End the call and contact the person or organisation through a number you already know and trust.
Can AI voice cloning be used legitimately?
Yes. With proper consent, voice synthesis can support dubbing, audio production and accessibility tools. It can also help people who expect to lose their speech to preserve a version of their own voice. The main disputes concern use without consent, fair payment, disclosure to audiences and deliberate deception.
Sources and further reading
- SoaplandTV: report on Derry Girls stars backing an AI voice campaign and on the petition’s signature total, used in summary form only
- Equity: UK performers’ union, with material on AI, consent and performers’ rights
- Intellectual Property Office and Information Commissioner’s Office: UK public bodies covering copyright policy and data protection
- National Cyber Security Centre: UK government guidance on phishing, social engineering and related fraud risks
Surfaced from the google:GB signal “AI voice campaign petition”. AI-assisted draft, editorially reviewed.
