Most deepfake programmes need only one picture to create a semi-believable facial model, and experts say AI can make voices indistinguishable from the real thing with just a few minutes of audio. This ...