When your eyes are no longer a witness: video and voice manufactured by a machine, of a person who never made them.
Cinema has had doubles for a century: the stuntman falls off the roof, the audience sees the star. Hundreds of people and months of work to fool the eye for two seconds. A deepfake is the same trick, except the double is a model, the set is a graphics card, and the whole crew fits into one line of text. What used to be an industry became a button.
The danger is not mainly in big politics, however much that gets discussed. It is in the everyday: the voice of “your daughter” urgently asking for money over the phone. “The boss” ordering a transfer over a video call. The scams are not new - what is new is how convincing the mask is.
What actually helps
Three habits. First: urgency plus money equals pause - call back yourself, on the number you know, not the one calling you. Second: read the context, not the pixels - who published the clip, does anyone else have it, what does the source who should know say. Third: there are technical markers too - watermarks and content-provenance standards mark what is generated, even if they are not everywhere yet.
Our read: the panic of “nothing can be trusted anymore” is exactly as wrong as naivety. Counterfeit banknotes have always existed - which is why banknotes carry protections and people have the habit of holding them up to the light. The same habit, carried over to video and voice, works here too.