‘20 minutes of terror’: AI boosts US voice impersonation scams - The Straits Times

In June 2026, Liz Benz, a 46-year-old insurance broker, was subjected to a 20-minute ordeal when she received a call from an unknown number featuring a voice that perfectly mimicked her 16-year-old son, Fred. The caller, sounding distressed and crying for help, claimed that Fred’s friend had been killed and that he was being held hostage, leading Benz to believe her son was in immediate danger.

What happened

The attack utilized AI voice cloning technology to replicate the tone, enunciation, and cadence of Benz’s son. The scammers instructed her to deliver cash to a nearby Walmart to secure his release. The deception was only broken when Benz received a selfie of her son smiling at a football game, confirming he was safe. Experts note that these scams often leverage small samples of a target's voice, sometimes only seconds long, harvested from social media or voicemails to create realistic replicas.

This incident is part of a broader trend of AI-enabled hoaxes. According to the FBI, Americans lost over US$893 million to such scams in 2025. Cybersecurity professionals emphasize that these attackers do not always require perfect voice replication; instead, their primary goal is to create sufficient emotional uncertainty and urgency to force the victim to act before verifying the situation. The scam often involves a scripted sequence where the primary "loved one" is joined by other voices posing as attorneys or bank tellers to heighten the pressure. Similar attacks have targeted individuals like Philadelphia attorney Gary Schildhorn, who also fell victim to a voice-cloned call claiming his son needed bail money. As these tools become more accessible, authorities and security experts are increasingly warning the public about the rising frequency of these sophisticated impersonation tactics.