1 in 3 People Got a Fake AI Voice Call Last Year

The numbers from Hiya's 2026 global report are blunt. Nearly 1 in 3 consumers worldwide received an AI deepfake phone call in the past year. In the U.S., the average person fields 10 unwanted calls per week, and 1 in 12 Americans lost money to a phone scam, with an average loss of $682. On top of those losses, 34% of consumers globally said their trust in phone calls declined over the past year.

That erosion has a second cost. People stop answering. They miss calls from doctors, schools, and family members in genuine emergencies. Scammers win twice: once when they take money, and again when they make everyone afraid to pick up.

Why Your Ears Can No Longer Save You

Voice cloning no longer requires sophisticated equipment or weeks of training data. Modern AI tools can replicate a person's voice from as little as 20 to 30 seconds of audio pulled from a voicemail, a short video, or a social media clip. Once cloned, that voice can say anything.

Human ability to detect synthetic audio is genuinely limited. Research puts accuracy at around 55 to 60 percent under controlled conditions, barely better than a coin flip, and that is before emotional pressure enters the picture. When the voice sounds like your child or your parent, asking yourself "does this sound like them?" is no longer a reliable test. Scammers design their attacks around exactly that vulnerability.

The classic grandparent scam has run for years on urgency and love: a caller claims to be a grandchild in crisis, asks for bail money or emergency funds, and pleads for secrecy. AI voice cloning made it worse. Now the caller doesn't just claim to be a grandchild. They sound like one. Families have wired thousands of dollars before thinking to call back on a known number. By then, the money is gone.

The problem is not gullibility. The problem is that the tool people have always used to verify identity, the voice itself, has been compromised.

A Check That Doesn't Rely on Sound

Calling back on a known number helps, though it takes presence of mind when you are scared, and a failed callback does not prove the original caller was legitimate. Security and fraud teams face the same pressure on the business side: AI-generated voices are already being used to impersonate executives and push through fraudulent wire transfers.

What individuals, families, and teams need is something simple, shared, and impossible for a scammer to know. Trust Onion is a free app that gives people a shared set of three rotating codewords, calculated locally on each device with no server required, working offline. When a call comes in from someone claiming to be a family member or a colleague, one question settles it: "What are the words?"

AI can clone a voice in seconds. It cannot know three codewords rotating on a private schedule that never touch a central server. For higher-stakes business requests, the same three-word challenge works as a practical second factor before approving a wire transfer or a vendor change.

The Hiya report makes clear that AI voice scams are widespread and already costing American consumers an average of $682 per incident. The voice might be convincing. The urgency might feel real. But the words are the one thing a cloned voice cannot borrow.

A controlled voice-cloning safety demonstration

Try our tool to see just how easy voice cloning is. This controlled safety demonstration exists only to show how easy voice cloning is and to make the risk tangible. It is not a real incident or proof of anyone's identity. Your submitted sample is handled transiently to generate the fixed demonstration audio. Trust Onion does not identify you or the person from the sample. Trust Onion does not store the submitted recording, clone, or generated audio.