Calls with "fake voices" are becoming increasingly realistic — now AI answers them. MegaFon's virtual assistant "Eva" has been updated and learned to recognize synthesized speech and deepfakes right during the conversation.
If the system notices suspicious signs, it immediately warns the user: the interlocutor may not be a human, but a neural network. And in case of obvious fraud, the call is automatically terminated.
"Eva" analyzes several sound parameters at once — from the structure of speech to the characteristics of the voice wave. This allows you to distinguish ordinary automatic calls (for example, from delivery services) from dangerous scenarios where attackers fake the voice of relatives, bosses, or famous people.
Protection against deepfakes was one of the most anticipated features — it was in the top 3 user requests.
Now "Eva" is turning not just into an assistant, but into a full-fledged security tool. The service already has several modes:
- "Secretary+" can answer calls and record conversations;
- "My Security+" blocks spam and suspicious numbers;
- "Family Security+" extends protection to loved ones.
In essence, this is another step towards AI not only simplifying life, but also protecting against new digital threats — especially those where distinguishing reality from forgery is becoming increasingly difficult.
