Beware the AI voice thieves

After years of blissful indifference, finally I’m scared of AI. I’ve been complacent, slept soundly beside my husband as he stares and mutters, sleepless with anxiety about robots. But now I’m frightened too. What happened was this.

A few weeks ago a friend received a phone call from her son, who lives in another part of the country. ‘Mum, I’ve had an accident,’ said the son’s voice. She could hear how upset he was. Her heart began to pound. ‘Are you OK? What happened?’ she said.

‘I’m so sorry Mum, it wasn’t my fault, I swear!’ The son explained that a lady driver had jumped a red light in front of him and he’d hit her. She was pregnant, he said, and he’d been arrested. Could she come up with the money needed for bail?

It was at this point that my friend, though scared and shocked, felt a prick of suspicion. ‘OK, but where are you? Which police station?’

The call cut off. Instead of waiting for her son to ring back, she dialled his mobile number: ‘Where are you being held?’ But this time she was speaking to her real son, not a fake of his voice generated by artificial intelligence. He was at home, all fine and dandy, and not in a police station at all.

Voice scams are on the rise in a dramatic way. They were the second most common con trick in America last year, and they’re headed for the top spot this year. In the UK: who knows? Our fraud squad doesn’t seem to be counting.

It’s just so horribly easy to con people when you’ve an AI as an accomplice. With only a snippet of someone speaking to learn from – a Facebook video, say – an AI tool can produce a voice clone that will sound just like them, and can be instructed to say anything.