Can AI Detect Deepfake Video Calls in Real Time?
Your CFO joins a Zoom call and asks the finance team to wire $25 million. The face looks right. The voice matches. Forty minutes later, the real CFO finds out nothing was scheduled. The Arup fraud in early 2024 unfolded exactly this way because detection did not save them. No Zoom plugin flagged the deepfake. No audio analyzer caught the clone.
The obvious question follows: can AI detect deepfake video calls in real time? And if the tools exist, why did Arup lose $25 million?