- The FBI's 2025 Internet Crime Report tracked AI-enabled scams as its own category for the first time: $893 million in total losses.
- Voice cloning tools can produce a convincing copy of someone's voice from a source clip as short as a few seconds, often pulled from something already posted publicly.
- The fix is not avoiding phones or social media. It's a verification habit: hang up and call back on a number you already have, every time, no exceptions for urgency.
The story is old: a panicked call, a relative in trouble, and a request for money right now, before you have time to think. What's new is what that call sounds like. In 2025, the FBI's Internet Crime Complaint Center (IC3) tracked artificial-intelligence-enabled fraud as its own category for the first time in the report's history, and tied more than $5 million in losses specifically to calls where a cloned voice played a family member in crisis.
What actually changed
The "grandchild in trouble" call isn't new. What's changed is the voice on the other end. Voice-cloning tools can now produce a convincing copy of someone's voice from a short source clip, and public reporting on real cases describes clips as brief as a few seconds pulled from a voicemail greeting or a video already posted online. The caller doesn't need to sound "close enough." They can sound exactly right.
That's a meaningful shift from the version of this scam that relied on a stranger's voice and a bad phone connection to explain away the difference. The IC3 report's 2025 numbers reflect that shift: $893 million in total AI-related fraud losses across 22,364 complaints, with $632 million of that tied to AI-driven investment scams and $352 million reported by adults age 60 and older. The distress-scam slice, cloned voices used specifically for family-emergency calls, accounted for more than $5 million on its own.
Why this beats the old version of the scam
The classic family-emergency scam had a built-in weakness: something usually sounded off. A voice that was almost right, a detail that didn't match, a connection that was too garbled to be sure. That weakness is what a cloned voice removes. If the voice sounds exactly like your relative, "it didn't sound quite like them" stops being a signal you can rely on.
That's why the defense has to move from listening harder to verifying differently.
The verification habit that actually works
None of these require assuming every call is fake. They just require one extra step before money moves:
- Hang up and call back. Use the number already saved in your phone for that person, never a number the caller gives you or a callback number in a text.
- Ask something only they'd know, ideally something you agreed on in advance rather than something guessable from social media (a childhood pet's name is often public; a private family reference usually isn't).
- Call another family member first if you can't reach the person directly. A real emergency involving one relative is often known to another.
- Treat urgency itself as the warning sign. Scammers need you to act before you verify. A genuine emergency almost always survives a two-minute pause.
See our family fraud protection kit for a printable version of this, including how to set up a family safe word in advance, before you ever need it.
If you've already sent money
Speed matters, but so does knowing what's actually recoverable. If the payment went by wire or a peer-to-peer app like Zelle, see our guide to Zelle and wire scam refunds for the legal distinction that determines whether your bank is required to refund it, and what to do in the first hour regardless.
Quick answers
How much has been lost to AI voice cloning scams? The FBI's 2025 Internet Crime Report ties $893 million in total losses to AI-enabled fraud, with more than $5 million specifically to cloned-voice family-emergency calls.
Can I tell a cloned voice from a real one? Not reliably, and that's the point of this article. Verify through a channel you control instead of trying to judge the voice itself.
What's the single best protection? A family safe word, agreed on before you need it, plus a habit of hanging up and calling back on a number you already have.
Sources
- FBI 2025 Internet Crime Report: the source for AI-scam loss figures cited above, the report's first year tracking AI-enabled fraud as its own category.
- FBI IC3: Elder Fraud: reporting channel and background on elder-targeted fraud schemes.
- FTC: Report Fraud: federal scam-reporting portal.
Methodology
Loss figures are drawn directly from the FBI's 2025 Internet Crime Report (IC3), the first edition to report AI-enabled fraud as a distinct category. This is general educational information, not personalized security advice; if you believe you've been targeted, report it to the FTC and, for financial losses, your bank and the FBI's IC3 directly.
What to Do Now
Frequently Asked Questions
How much has been lost to AI voice cloning scams?
How much audio does it take to clone someone's voice?
What's the fastest way to check if a family-emergency call is real?
Should I stop posting videos of myself online to avoid this?
The 5-minute money briefing
One email per week. New rates, fed moves, and what to actually do about them.
No spam. Unsubscribe anytime.
Compare today's top savings options, ranked by SwitchWize's composite score: rate, trust, and switching ease.
Editorial review
What changed since the last update
Was this guide helpful?
Found an inaccurate, outdated, or missing claim? Report a correction. We verify reports against the relevant source before changing a guide or ranking.
