Deepfake
AI-generated video, audio, or images that convincingly depict real people saying or doing things they never said or did, increasingly used to enable fraud and disinformation.
Also known as: synthetic media, AI-generated video
Last reviewed: 1 June 2026
Deepfakes use deep learning (a form of artificial intelligence) to synthesise realistic media by superimposing a person's likeness onto another's body, or generating new video from photos and voice recordings. Early deepfakes were detectable by careful inspection, but quality has improved dramatically.
In the fraud context, deepfakes are used to impersonate executives in video calls to authorise wire transfers, create fake celebrity endorsements for investment scams, produce non-consensual intimate imagery for sextortion, and fabricate 'evidence' of events that never occurred.
Signs of a deepfake include unnatural blinking patterns, inconsistent lighting or shadows around the face, audio that is slightly out of sync with lip movements, and blurring at the edges of the face. Verification by calling the person independently on a known number remains the most reliable defence when video authenticity matters.