6 minSociety
AI voice cloning scams are fooling families, with a Martinez mother wiring $5,400 after a fake kidnapping call
Criminals are using AI to clone voices from short audio clips, and experts say families should agree on a safe word to verify emergency calls.
Artificial intelligence can now produce a convincing copy of a person's voice from as little as three seconds of speech, and criminals are using that capability to convince relatives that a loved one is in danger. In tests by the cybersecurity firm McAfee, a three-second clip was enough to generate a clone with an 85 percent match to the original speaker. Experts say the technique is increasingly showing up in fraud, and that a familiar voice can no longer be treated as proof of identity.
The scale of the problem is significant. A survey by Gallup and the Stop Scams Alliance estimated that about 15 million American adults were scammed out of money in 2025, losing roughly $68 billion. The full report found that 12 percent of the scams reported involved artificial intelligence or a deepfake. Researchers say that figure may undercount the problem, because victims do not always realize AI was used. Scammers can collect voice samples from videos people casually post on platforms such as Instagram and TikTok, then use the clones to pressure family members into sending money.
One Martinez, California, mother described how quickly the deception unfolded. Deborah Del Mastro received a call from an unknown number one morning in May. The caller claimed a Mexican drug cartel had kidnapped her 37-year-old daughter, then played what sounded like her daughter crying for help. « It was my daughter's voice having an absolute panic attack, scared, telling me 'I love you, mom, I'm so sorry, I'm so scared' and then they just cut it off, » Del Mastro told ABC7 News. She said the caller gave rapid orders, telling her not to speak, to get dressed and leave, and to act immediately.
Believing her daughter's life was at risk, Del Mastro followed those instructions for about five hours. She wired $5,400 to Mexico from multiple locations and was eventually told her daughter would be released at a grocery store. When she arrived and could not find her, she called her daughter directly and learned that she had been at work the entire time. Martinez police are investigating, but Del Mastro told ABC7 she does not expect to recover the money.
Kimberly Hyun, an assistant professor of marketing at the University of Cincinnati's Carl H. Lindner College of Business, studies how a person's voice shapes persuasion and consumer choice. She told Fortune that these scams work because of how the brain processes sound. Her research finds that when a voice's timbre resembles someone we know, people tend to lower their defenses and become more inclined to believe what they hear. A family member's voice, she said, « can actually trick you through trust. »
The fraud is not limited to older people. Gallup found no difference in scam victimization rates by age, and experienced professionals have also been deceived. In 2024, a finance worker at a multinational firm in Hong Kong was tricked into paying out HK$200 million, or about $25.6 million, according to Hong Kong police. The fraudsters used deepfakes to pose as the company's chief financial officer on a video call. The worker had initially suspected a phishing attempt, but set his doubts aside because the other participants looked and sounded like colleagues he recognized. All of them were fakes. The company was later identified as the British engineering firm Arup.
The tools behind these scams are widely accessible. In March 2025, Consumer Reports assessed six popular AI voice cloning products and found that four of them, ElevenLabs, Speechify, PlayHT and Lovo, allowed researchers to clone a voice easily from publicly available audio. Those four companies used no technical mechanism to confirm the speaker's consent; users only had to check a box attesting they had the legal right to clone the voice. Two others, Descript and Resemble AI, had taken steps to make nonconsensual cloning harder. Grace Gedye, a policy analyst at Consumer Reports, said, « Our assessment shows that there are basic steps companies can take to make it harder to clone someone's voice without their knowledge, but some companies aren't taking them. »
Because a familiar voice is no longer reliable proof of identity, Hyun recommends that families agree on a code word that only they know. « You really need to create a safe code with your loved ones. Make sure that they're real, » she said. Beyond that, people should be wary of calls from unknown numbers and should not stay on the line if someone claims a relative is in trouble. « You just need to hang up and just call your family member again, » Hyun advised. Victims can report fraud to the Federal Trade Commission at ReportFraud.ftc.gov and to the FBI's Internet Crime Complaint Center. In the Gallup survey, only 13 percent of scams were reported to the FTC or federal law enforcement.
4
