BBB: AI Scam Complaints Hit 100,000 as Voice Cloning Explodes

The Scam That Sounds Like Family

Your phone rings. It's your son's voice. He's panicked, says he's been in an accident, needs money fast, please don't tell anyone. The voice is right. The emotion sounds real. You wire the money.

Then your actual son calls an hour later.

This isn't hypothetical. The BBB is warning consumers across the country that AI-powered voice cloning has become a primary tool in automated fraud. Scammers need just a few seconds of publicly available audio, a voicemail greeting, a social media video, a clip from a podcast, to build a convincing clone of almost anyone's voice.

How the Automation Works

What makes 2026 different from earlier waves of phone fraud is scale. AI doesn't sleep, doesn't get tired, and doesn't need to hire a call center.

According to the BBB's warning issued in August 2026, scammers now combine multiple AI tools into automated pipelines. A voice clone handles the call. A phishing message handles the follow-up. A deepfake video handles the proof. Each piece reinforces the others, and none of it requires a human scammer on the other end in real time.

The BBB tracks three main AI-assisted methods.

Voice cloning calls. A cloned voice impersonates a family member, friend, or company executive. The caller creates urgency, requests money or sensitive information, and discourages the target from hanging up or calling anyone else.

AI-generated phishing messages. Text and email scams that used to be easy to spot because of bad grammar are now polished and personalized. AI writes them at scale, tailored to the recipient.

Deepfake video. For higher-stakes fraud, scammers use video calls where the person on screen looks and sounds like someone the victim trusts. In February 2024, engineering firm Arup lost $25 million when an employee authorized a wire transfer after joining a video call where every participant, including someone who appeared to be a senior executive, was a deepfake.

Why Voice Cloning Is the Hardest to Catch

Of all the AI fraud methods, voice cloning catches people off guard most often. We're trained from childhood to trust the voices of the people we love. That trust runs deep and it runs fast.

Voice cloning now requires just 20 to 30 seconds of audio, shorter than most voicemail greetings. Anyone who has posted a video online, spoken at a public event, or left a voicemail has probably given scammers enough to work with.

Human detection accuracy for cloned voices sits at around 55 to 60 percent, according to researchers. That's barely better than a coin flip. If you can't reliably tell a fake voice from a real one by listening, you need a different method entirely.

The "Are You Sure It's Really Them?" Problem

Most families don't have a plan for this. When a call comes in sounding like someone you care about, your instinct is to help, not to interrogate. Scammers know this and script the call to keep you emotionally activated and moving fast.

The BBB's advice is to pause, call back on a known number, and verify. That's sound advice, but it assumes the person on the other end will let you hang up, that you'll remember to do it in a moment of panic, and that you have a fast, reliable way to confirm identity when seconds feel like they matter.

What Actually Stops a Voice Clone Cold

AI can clone a voice. It cannot know something it was never told.

That's where a family's three rotating codewords come in. Trust Onion is a free app that gives every family a set of three words that rotate every 60 seconds, calculated locally on each person's phone. No server. Works offline. If someone calls claiming to be your daughter, you ask one question: "What are the words?"

A cloned voice can't answer that. A deepfake can't answer that. A scammer who has studied every public video of your family member for a year still can't answer that, because the words changed 40 seconds ago and will change again in 20 more.

The words rotate on a shared schedule, so everyone in the family sees the same three words at the same time. If the caller knows the words, it's really them. If they hesitate, deflect, or give the wrong answer, hang up.

For situations where more certainty matters, Trust Onion also lets family members send a "Proofie": a selfie with the current three words overlaid and signed, so the recipient can verify it matches. That combines what you look like, what you know, and when you sent it. No AI can fake all three at once.

This is exactly the kind of call those 100,000 BBB complaints describe, and it's the kind of call three rotating codewords stop in under ten seconds.

What You Can Do Right Now

The BBB recommends a few practical steps alongside any technical protection.

• Slow down. Urgency is a weapon. If someone is pressuring you to act immediately, that's the scam at work.
• Call back on a number you know. Don't use a number the caller gives you. Look it up yourself.
• Talk to your family now. Before a call comes, agree on how you'll verify each other. A shared codeword or verification phrase, set up before you need it, is the fastest way to confirm identity under pressure.
• Report it. The BBB's Scam Tracker at bbb.org/scamtracker accepts reports and helps others recognize patterns.

The complaints aren't slowing down and the technology isn't going away. Families who talk about this ahead of time and set up something simple to verify each other are the ones who hang up before the money leaves their account.