It’s the new badge of celebrity status that no one wants. Jennifer Aniston, Oprah Winfrey and Kylie Jenner have had their voices cloned by scammers. Online blaggers used artificial intelligence to spoof the Tiggerish pitch of TV financial advisor Martin Lewis. And this weekend, David Attenborough described himself as “deeply disturbed” to have discovered that his cloned voice had been used to deliver American partisan news bulletins.
Now experts have warned that voice cloning is outlawing the law as technologists refine previously clunky speech generators into models capable of emulating the subtlest pauses and breaths of human intonation.
Dr Dominic Lees, an expert on AI in film and television who advises a UK parliamentary committee, told The Guardian on Monday: “Our privacy and copyright laws are not up to date with what this new technology presents , so there is very little that David Attenborough can do.”
Lees is advising the House of Commons culture, media and sport select committee on an inquiry that will look into the ethical use of AI in film-making. It also convenes the Synthetic Media Research Network, whose members include the firm that is making an artificially intelligent version of the late chat show interviewer Michael Parkinson, which will result in an eight-part unscripted series, Virtually Parkinson, with new guests. That voice cloning project is being done with the consent of Parkinson’s family and estate.
“The government definitely needs to consider (voice cloning), because it is a major fraud problem,” Lee said. “You need the stick of government regulation to deter (misuse)… we can’t let it become a pitched battle.”
AI voice cloning scams increased by 30% in the UK last year, according to investigation by NatWest Bank this month. Another lender, Starling Bank, found that 28% of people had been targeted by an AI voice cloning scam at least once in the last year.
Scammers also reportedly use voice cloning to perpetrate a version of the “hi mom” text scam, in which scammers pose as children who urgently need their parents to send funds. On already confusing phone lines, it can be difficult to detect that a supplicant child is the clone of a scammer. Consumers are advised to check by hanging up and calling a trusted number again.
People whose voices are cloned without their consent consider it more than a nuisance. Attenborough told the BBC on Sunday: “Having spent my whole life trying to tell what I believe to be the truth, I am deeply disturbed to discover that these days others are stealing my identity and I am very much opposed to the use to say what you want.”
When a new voice option in OpenAI’s latest AI model, ChatGPT-4o, featured tones very close to those of actress Scarlett Johansson, she said she was shocked and angry because the voice “sounded so eerily similar to mine.” that my closest friends and the media couldn’t tell the difference.”
The rise of cloned voices raises the question of what is lost about real human tones. Lees said: “The big problem is that AI doesn’t understand emotions and how that changes the way a word or phrase can have an emotional impact, and how the voice is varied to represent that.”
The voice-over industry, which provides voices for advertisements, animations and instructional training, has to respond quickly to technological advances. Joe Lewis, head of audio at London’s Voiceover Gallery, which has provided real human voices for adverts for Specsavers and National Express, said it had already cloned the voices of some of its artists.
He said the AI seemed to work better with male English voices, perhaps because that reflected bias in the type of recordings that had been used to train the algorithm, but cautioned that overall “there’s something about the way it’s generated that It makes you feel better.” less attentive.”
“When the AI (voice) breathes, it is a very repetitive breathing,” he said. “The breaths are in the right place, but they don’t feel natural… (But) can it get to the point where it’s really perfect? “I don’t see why not, but getting to the full emotional spectrum is a long way off.”