In 2024 criminals tried to trick staff at WPP, the world's biggest advertising group, using a cloned voice of its chief executive and a fake video meeting. The attempt failed, but it showed how easy these attacks have become.
Attackers rarely fake everything. They mix real and AI content because it's harder to spot. Security teams now see:
a cloned voice calling the IT help desk to reset a password or add a new device
a deepfake face of an executive on a video call, driven by a real person speaking
real footage with a cloned voice added, so someone appears to say something they never said
a real person on a call reading answers written by AI
A tool that checks only the face, or only the voice, will miss the part that's fake. Good detection checks every layer of the same content:
the video, frame by frame, for signs of a deepfake face
the audio, for signs of a cloned or synthetic voice
the words spoken, turned into text and checked for AI writing
AI Aware checks all three together and shows which parts are real and which are AI. It is over 97% accurate on text, images, video and audio, and was built with computer scientists at City St George's, University of London. Read more on the deepfake and AI detection for cybersecurity page.