It’s the call that all financial directors now dread. At the other end of the line, the voice is calm, familiar, perfectly recognizable: that of the CEO. The order is urgent, confidential, and concerns a strategic acquisition requiring an immediate transfer of several million euros. The problem? It is not the boss speaking, but an artificial intelligence trained from a few minutes of their public interventions.
The Era of Social Engineering 2.0
The phenomenon of audio deepfakes, or voice cloning, has left the realm of research laboratories to become a major operational threat. While classic “CEO fraud” relied on intimidation by email, AI today brings almost infallible technical credibility. In 2024, less than thirty seconds of sound sample, retrieved from YouTube or during a webinar, is enough to create a vocal model capable of holding a conversation in real time.
The figures are alarming. According to the latest cybersecurity reports, AI-based attack attempts have jumped by more than 3000% in one year. The most emblematic case remains that of a multinational in Hong Kong, where an employee transferred 25 million dollars after a video conference where all participants, except him, were deepfakes.
An Accessible and Inexpensive Technology
What particularly worries the experts at 24h00.info is the democratization of tools. No need to be an elite hacker to clone a voice. Legitimate SaaS (Software as a Service) platforms, designed for dubbing or content creation, are hijacked by cybercriminal groups. For a few dozen euros per month, anyone can access voice synthesis algorithms capable of reproducing intonations, language tics, and even regional accents.
How are Companies Organizing?
Faced with this invisible threat, the technological response is no longer enough. Firewalls and antivirus software are powerless against a convincing phone call. Companies must now rethink their human validation protocols. Several strategies are emerging to counter these attacks:
- The Security Password: The establishment of a verbal secret code, known only to members of management and financial services, to validate any sensitive transaction.
- The Call-back Procedure: The obligation to systematically call the requester back on a different communication channel (internal line or encrypted messaging).
- Behavioral Training: Making employees aware that a voice, even a familiar one, is no longer sufficient proof of identity.
- Biometric Analysis: The deployment of software solutions capable of detecting artificial frequencies specific to AI-generated voices.
The Future of Digital Trust
The question is no longer whether a company will be targeted, but when. The transition from text to sound, and soon to real-time video, marks a technological breakthrough. Trust, a pillar of business relationships, becomes a vulnerability. For technical departments, the challenge is immense: it is about securing not only networks but human interaction itself.
As GenAI tools progress, the only effective barrier remains increased vigilance and systematic mistrust of exceptional requests, even when they seem to come from the highest-ranking person in the organization.
