Humans cannot detect deepfake speech 27% the time, researchers from University College London (UCL) have found during a recent study. The study presented 529 individuals with genuine and deepfake audio samples and asked them to identify the deepfakes.

Source: Infosecurity