Entering into a meeting with your boss and several familiar coworkers inside a video conference is the next area vulnerable to cybercrimes, and it’s all because of deepfake technology. Researchers at the Fraunhofer Institute for Secure Information Technology SIT are working on a real-time warning system designed to identify attempted fraud during corporate video conferences.
The report states that criminals are increasingly targeting video meetings for identity theft and financial scams, taking advantage of the trust people place in familiar faces and voices. The intention is to flag suspicious activity while the meeting is still taking place. This gives an employee a chance to stop before following an expensive instruction from an AI-generated executive.
How an entire meeting is being faked now
Honor Deepfake video detection Honor
The threat isn’t as simple to spot now. There’s no uncanny valley vibe os a single poorly synchronized face. Researchers pointed to a case in Hong Kong where an employee joined a videoconference containing several apparent colleagues. Everyone else in the meeting was fake. The criminals had reportedly accessed internal video material and used AI-generated voices to impersonate company employees, eventually persuading the victim to transfer almost 24 million euros. He discovered the fraud only after speaking to his real boss afterwards.
Human judgment offers limited protection against a convincing setup. A separate 2025 study showed students a manipulated video impersonating their professor using an openly available real-time face-swapping tool. Twenty-five of the 34 participants, or 74%, failed to recognize that the video had been altered before researchers told them. Once several familiar faces appear together, their presence can reinforce the illusion and make an unusual request look legitimate.
A warning should start the verification process
Fraunhofer’s existing media-forensics research combines deep learning with signal processing to search for clues such as blurred facial structures, inconsistencies between image regions, and other minute oddities. There is still reason to treat any warning carefully.
One researcher, Martin Steinebach, has previously cautioned that automatic deepfake detectors can produce enough errors to cause confusion when used during live calls. So the system being built by Fraunhofer could provide the crucial nudge before urgency and a familiar AI-generated face convince someone to act.