A deepfake is AI-generated or AI-manipulated audio, image or video that passes itself off as real – typically a real person's face or voice inserted into something that person never said or did.
The technique behind it is the same one that drives the creative image tools – the difference is the intent. A deepfake is designed to deceive: a CEO asking for an urgent transfer, a politician saying something scandalous, a family member in distress. The quality is now so high that the eye can rarely tell.
The answer is not primarily technical but organisational: fixed verification routes for sensitive decisions, agreed channels for critical communication – and a culture in which it is legitimate to call back and check.
In practice
Several Danish companies have already experienced CEO fraud using cloned voices. The simplest countermeasure is low-tech: no transfers or confidential information on the strength of a single request – however genuine it seems.
How to explain it to leadership
»Deepfakes do not mean that everything is fake – they mean that authenticity has become something we must be able to prove. That makes the sender's responsibility greater, not smaller.«
Trust and authorship are also at the heart of the article When creativity loses its sender.