Scammers can clone a voice from just three seconds of audio, warns Spanish cybersecurity agency
Spain's national cybersecurity agency says AI can now reproduce a person's voice using only three seconds of publicly posted audio, enabling new phone scams.
INCIBE published new advice on September 14 about a surge in AI-driven voice-cloning scams, where criminals can recreate a relative's voice from merely three seconds of audio found on social media. The technology now mimics accents, pauses and breathing, allowing scammers to respond dynamically rather than replaying a fixed script. In 2026, more than 135 million suspicious calls were blocked in Spain, with roughly one-third involving synthetic voices, and victims typically lose about €577 each.
The agency stresses that callers should end the conversation if pressure builds, verify the caller by dialing a saved number, and establish a unique verification phrase in advance. A free helpline (017) is available for doubts or reporting, and future measures may include AI-generated audio labeling and phone-app verification features.
Why it matters
AI voice cloning makes phone fraud easier, putting families at risk of losing money.
In this story
