Generative AI models (like ElevenLabs or VALL-E) can analyze a 3-second sample of a person's voice and generate infinite speech in that voice. This enables targeted "Vishing" (Voice Phishing) attacks that bypass even the most suspicious victims.
The Kidnapping Scam
A parent receives a call. They hear their child screaming for help. "Mom, they have me!"
It sounds exactly like them.
The "Kidnapper" demands ransom.
In reality, the child is safe at school. The AI cloned their voice from a TikTok video.
1. Video Deepfakes
Real-time video face-swapping is now possible in Zoom calls.
Hackers interview for remote IT jobs using Deepfakes.
Once hired, they gain access to the company network and deploy ransomware.
2. Detection
It is becoming harder to detect.
Current signs:
- Unnatural blinking.
- Lip-sync delay.
- Audio artifacts (robotic glitches).
The best defense for families/companies is a "Safe Word". If someone calls acting strange, ask for the safe word.