ROR Labs cover: $893M first count. The FBI opened an AI-enabled fraud section for the first time in 2025: 22,364 complaints. Your voice is no longer proof that it is you.
|

Your Voice Is No Longer Proof That It Is You

Further reading: Cultish — Amanda Montell (Harper Wave, 2021). No trade book yet covers AI voice-cloning or pig-butchering scams specifically; this is the closest available account of how persuasive, manipulative language actually works on a target. Find it on Amazon (paid link)

Disclosure: this article contains affiliate links, marked (paid link). If you buy through one we may earn a commission, at no extra cost to you. As an Amazon Associate I earn from qualifying purchases. It costs you nothing and it does not change what we recommend.

Key takeaways · 8 min read

  • Reported losses to cyber-enabled crime reached about $21 billion in 2025, from over a million complaints.
  • The FBI’s first AI-fraud category recorded 22,364 complaints and roughly $893 million.
  • Adults over 60 lost about $7.7 billion, up 37%, because they hold assets and are targeted, not because of poor judgement.
  • A convincing voice clone needs only seconds of public audio.

For most of human history, recognising someone’s voice was reliable evidence that you were talking to them. That stopped being true recently enough that most people have not adjusted, and this article is about making that adjustment.

A few seconds of audio from a video posted to social media, a voicemail greeting or a work presentation is now enough to produce a convincing imitation of a specific person saying something they never said. In 2025 the FBI logged its first dedicated category for this.

A telephone seen from above on a table with the handset lying beside it, next to a mug and a notepad.
It sounded exactly like them. It was never them.

The numbers, and who is losing

FBI Internet Crime Report, 2025

Released 6 April 2026. Over one million complaints, up from 859,532 the year before.

$21Btotal reported losses to cyber-enabled crime in 2025
$893Min the report’s first-ever AI-enabled fraud section — 22,364 complaints
AI fraud is still a small slice of the total. It is the growth rate and the mechanism that make it worth a separate article — this is the first year the category existed at all.

Source: Federal Bureau of Investigation, “Cryptocurrency and AI Scams Bilk Americans of Billions,” IC3 2025 Internet Crime Report, 6 April 2026.

Where the losses landed

By age group and category, 2025.

Adults 60 and over$7.7B
  — change on 2024+37%
People over 60 lost about $7.7 billion, the steepest year-on-year rise of any age group. Phishing and spoofing, extortion and investment schemes were the most frequently reported complaint types.

Source: FBI IC3 2025 Internet Crime Report (April 2026).

The older-adults figure needs care, because it is usually reported in a way that implies gullibility. The likelier explanations are duller and more actionable. Older adults hold more assets and are more often home to answer a phone, and they are targeted deliberately for both reasons. Being defrauded is not evidence of poor judgement when the attack is built specifically to defeat good judgement.

How the voice scam is built

A wall of dark picture frames of different sizes hung slightly askew.
Three seconds of you is already public.

The structure has been the same for decades. It used to be called the grandparent scam, and it worked with a stranger and a bad phone line. AI has changed one thing: the voice now matches.

The four moves, in order

Recognising the shape is more durable than recognising any particular script.

1. A VOICE YOU KNOWCloned from audio posted publicly. Distressed, often crying — which conveniently explains why the voice is not quite right.
2. AN EMERGENCYAn accident, an arrest, a hospital. Urgency is the active ingredient: it exists to stop you thinking, not to describe a real event.
3. SECRECY“Don’t tell Mum.” “My lawyer says not to discuss it.” This isolates you from the one thing that breaks the scam — another person.
4. AN IRREVERSIBLE PAYMENTWire transfer, cryptocurrency, or gift card numbers read aloud. These are chosen because the money cannot be recovered. This step is the tell.
No real emergency has ever required gift cards. If the story ends in a payment method that cannot be reversed, the story was written to end there.

Source: U.S. Federal Trade Commission, “Scammers use AI to enhance their family emergency schemes” consumer alert; FBI IC3 reporting.

The FTC’s advice is one sentence long and defeats the entire category: do not trust the voice; call the person back on a number you already have. A cloned voice cannot answer a phone you dial. That is the whole defence, and it costs nothing.

Why calling back works when listening does not

The attacker controls the inbound call. They do not control a number you already have.

THEY CALL YOU — every element is chosen by them caller ID (spoofable) the voice (cloned) the story (scripted) the urgency nothing here is evidence YOU CALL THEM — on a number already in your phone saved contact the actual person who is fine, and confused
If they insist you must not hang up, that is the scam identifying itself. A real emergency survives a five-minute call back.

Source: FTC consumer guidance on family emergency scams.

One case: a call from a “grandson”

One reported case comes from Good Morning America on 16 May 2023, told by a family. That the voice was AI-generated is the family’s account as reported.

The call. A caller, sounding distressed and made to sound like his grandson, phoned an elderly man. The caller said the grandson was in Mexico, had been drinking, had had an accident and needed money. The grandfather sent $1,000.

The unmasking. The grandson’s parents then called him directly. He was a college student in Dallas, and he was fine. The article does not say whether the money was recovered or whether police were involved.

What the Federal Trade Commission (FTC) advises. Resist the pressure: hang up, or say you will call right back. Ask a question only the real person would know. Contact the family member on a number you know is correct. Call another relative or friend, even if told to keep it secret. Be wary of demands for wire transfers, crypto, payment apps or gift cards. The FTC’s page says scammers use voice cloning to make such requests more believable.

One extra habit. The news article suggests a family code word and keeping social media private. The FTC pages we read do not mention a code word, so treat that as a family decision and not an official rule.

What we could not confirm. We could not verify that the voice was cloned rather than imitated, and one case does not tell us how common voice cloning is.

What actually protects you

All of it is free, and most of it is a conversation

There is no product for this. There is a habit.

1. AGREE A FAMILY CODE WORDOne word, agreed in person, never written down or sent by message. Anyone claiming an emergency has to say it. Simple enough for a child, and it defeats a perfect voice clone.
2. HANG UP AND CALL BACKAlways on a number you already have. Never a number the caller gives you, however official it sounds.
3. TREAT URGENCY AS THE SIGNALReal institutions and real relatives can wait ten minutes. “Right now, and don’t tell anyone” is the shape of the attack, whatever the story.
4. KNOW THE PAYMENT TELLSGift cards, wire transfers, crypto, or a courier collecting cash. No legitimate organisation asks for any of these. Ever.
5. HAVE THE CONVERSATION NOWWith parents, and with anyone who might get the call. Deciding the rule in advance is what makes it survive the adrenaline.
6. THIN THE PUBLIC AUDIOMarginal, and worth knowing: voice clones are trained on public video and voicemail greetings. It will not stop a determined attacker.

Sources: FTC consumer alerts on AI family-emergency scams; FBI IC3 2025 report.

The small amount of hardware that helps

Nothing here detects a fake voice. These reduce how often the call reaches someone in the first place.

CALL-BLOCKING PHONEFor a landline. Screening unknown numbers before they ring is unglamorous and effective — the scam cannot start if the call is not answered.
Browse on Amazon →
LARGE-BUTTON PHONE WITH PHOTO KEYSProgrammed one-touch keys to family. It makes calling back on a known number the path of least resistance, which is the entire defence.
Browse on Amazon →

We are not linking “AI deepfake detector” apps or identity-theft monitoring subscriptions. Consumer voice-detection tools are unreliable and improving far more slowly than the generators they are chasing — trusting one could make you more confident about a fake. Identity monitoring tells you after your data has leaked; it does nothing about a phone call. A code word costs nothing and works better than both.

Two mugs on a kitchen table in daylight, one of them untouched and still steaming.
The call that settles it is the one you make.

What to do

Steps sorted by when they matter. Each one says how strong the evidence is: Measured Official advice Untested

Before

  • Agree a family code word in person. Never write it down or send it by message. Official advice see Sources
  • Have the conversation now, with parents and anyone who might get the call. A rule decided in advance survives the adrenaline. Official advice see Sources

If it happens

  • Hang up and call back on a number you already have, never one the caller gives you. Official advice see Sources
  • Treat urgency as the signal. “Right now, and don’t tell anyone” is the shape of the attack. Official advice see Sources

What not to do

  • Do not pay by gift card, wire transfer, crypto or cash handed to a courier. No legitimate organisation asks for these. Official advice see Sources
  • Do not trust a deepfake detector app. Consumer voice-detection tools are unreliable. Untested see Sources

Step-by-step guide, with a printable checklist: Sent money to a scammer? What to do, in order

Questions people ask

Can I tell a cloned voice by listening?

Increasingly no, and the scam is designed around that. Distress is chosen partly because crying explains away any oddness in the audio. A defence built on your own ear rests on the one thing the attacker has already optimised against.

What if I have already paid?

Act immediately, because a small window sometimes exists. Call your bank, since wire transfers can occasionally be recalled within hours. For gift cards, call the issuer with the card numbers; funds are sometimes frozen if reported fast. Report to the FBI’s IC3 at ic3.gov and to the FTC at reportfraud.ftc.gov. And tell someone. Shame is the reason most of this goes unreported, and unreported fraud lets the same operation keep running.

Is my bank going to reimburse me?

It depends heavily on jurisdiction and on whether you authorised the payment yourself. Fraud where someone takes your money without permission is usually covered; fraud where you were persuaded to send it often is not, though rules in several countries are shifting toward reimbursement. Report it anyway and in writing, since the outcome frequently depends on how quickly and how formally it was raised.

My parent will not use a code word. It feels ridiculous.

Frame it as protecting you rather than them. People accept a rule for someone else’s benefit much more readily than one that implies they might be fooled, so “If someone rings you pretending to be me, I want you to have a way to check” lands very differently from “you might fall for this.”

The short version

  • Reported losses to cyber-enabled crime reached about $21 billion in 2025, from over a million complaints.
  • The FBI’s first AI-fraud category recorded 22,364 complaints and roughly $893 million.
  • Adults over 60 lost about $7.7 billion, up 37%, because they hold assets and are targeted, not because of poor judgement.
  • A convincing voice clone needs only seconds of public audio.
  • The shape is always: known voice → emergency → secrecy → irreversible payment.
  • Gift cards, wire transfers and crypto are the tell. No legitimate emergency ends with them.
  • The defence is hang up and call back on a number you already have. A clone cannot answer your call.
  • Agree a family code word in person, today. It is free and it beats every detection product on the market.

If you or someone you know has been targeted, report it to IC3 at ic3.gov and the FTC at reportfraud.ftc.gov. Reporting quickly occasionally allows funds to be stopped, and it is how these operations get traced.

On the links above: some are affiliate links, marked (paid link). If you buy through one we may earn a commission at no additional cost to you. As an Amazon Associate I earn from qualifying purchases. We link to product searches rather than specific items so recommendations do not break as models change, and we say plainly when we are choosing not to link something. Full policy: Affiliate Disclosure.

Sources

  • Federal Bureau of Investigation. “Cryptocurrency and AI Scams Bilk Americans of Billions.” IC3 2025 Internet Crime Report, 6 April 2026.
  • Federal Trade Commission. “Scammers use AI to enhance their family emergency schemes.” Consumer alert.
  • FBI Internet Crime Complaint Center (ic3.gov); FTC report fraud portal (reportfraud.ftc.gov).
  • Good Morning America. Father warns others about AI voice scams after his family lost $1k, 16 May 2023. goodmorningamerica.com.
  • FTC. Scammers use fake emergencies to steal your money. consumer.ftc.gov.

Similar Posts