News
HuffPost UK cites CRIF data showing AI voice cloning fear has tripled among Brits, and BBC figures putting 2025 fraud losses at £1.3bn.
Reported by
HuffPost UKAccording to HuffPost UK, a call from a loved one asking for money urgently is no longer proof that a loved one is on the line. Artificial intelligence can now clone a voice from clips found on social media, and criminals are using that to fake the one signal people have always trusted on the phone.
The report cites Europe-wide research from CRIF showing three quarters of Brits are worried about AI-powered fraud, up from just 24% a year earlier. Two thirds now believe fraud and cybercrime are a bigger personal threat than before. The BBC reported more than four million fraud cases in 2025, totalling £1.3bn stolen.
The scam does not depend on a convincing voice alone. It depends on speed. The fake call is built to stop a person thinking, not to survive scrutiny. That is why the red flags named by Central Bank all point the same way: pressure to act quickly, an unusual request that feels out of character, and a refusal to let the call end so the person can check.
That last one is the tell that matters most. A cloned voice can sound right. It cannot easily keep someone on the line while also letting them hang up and call back. The moment a caller resists being put down and redialled, the voice stops being the evidence and the behaviour becomes the evidence instead.
This is also why impersonation fraud, criminals posing as a bank, the police or an organisation, has actually fallen according to UK Finance, even as AI-enabled scams rise. People have been trained to distrust official-sounding callers. They have not yet been trained to distrust a call that sounds like their own child or partner. The scam has simply moved to the one voice nobody was told to question.
Hanging up breaks it, every time, because it removes the one thing the scam needs: a live, unbroken line of pressure. Calling the person back on their known number, or through another platform they actually use, does the same thing from the other side. A prearranged code word known only within a family works because it cannot be cloned from a social media clip, however good the voice sample is.
None of this requires spotting a fake voice. It requires refusing to make a decision while someone else is controlling the clock.