Deepfake Voice Cloning Scams: How They Work

How AI voice cloning scams work, who they target, and how to set up a family code word that stops them -- with the FBI's own advice.

Table of Contents

You answer the phone and hear your daughter’s voice. She is crying. She says she has been in a car accident and needs money for bail. The voice is unmistakably hers – the same pitch, the same inflection, the same way she says “Mom” when she is frightened. Except it is not her. It is a synthetic voice generated by artificial intelligence, and the person on the other end of the line is a scammer who wants your money. This article is part of our Phishing & Social Engineering guide.

Voice cloning scams represent one of the most psychologically devastating forms of social engineering to emerge in recent years. In December 2024 the FBI warned that criminals use generative AI to create short audio clips of a loved one’s voice to fake an emergency or a ransom demand, and the technology keeps getting cheaper, faster, and more accessible.

How Voice Cloning Technology Works

Modern AI voice cloning requires surprisingly little input to produce a convincing replica of someone’s voice. In a 2023 test, McAfee’s researchers produced an 85% voice match from just three seconds of audio. That audio can come from:

  • Social media videos – A TikTok, Instagram Reel, or YouTube video where the target speaks clearly.
  • Voicemail greetings – Calling someone’s phone and recording their outgoing message.
  • Public recordings – Podcasts, conference presentations, interviews, webinars, earnings calls.
  • Leaked audio – Voice messages from compromised accounts or data breaches.

The AI analyzes the audio sample and creates a voice model that captures the speaker’s unique characteristics – pitch, cadence, accent, speech patterns, and emotional inflections. Once the model is built, the attacker can type any text and have it spoken in the cloned voice in real time.

The quality of these systems has crossed the threshold of human perception. In controlled tests, people are unable to reliably distinguish between genuine recordings and high-quality voice clones. The technology is not perfect – there are occasionally subtle artifacts – but in the high-stress context of an emergency phone call, those artifacts are easily overlooked.

Common Voice Cloning Scam Scenarios

The Emergency Family Call

The most emotionally devastating variant. An attacker clones a family member’s voice and calls the target with an urgent emergency:

  • “Mom, I’ve been in an accident. I need you to send money right now.”
  • “Dad, I’ve been arrested. Don’t tell anyone. I need bail money.”
  • “Grandma, I’m in the hospital. Can you wire me money for the bills?”

The call creates panic, and the urgency overrides critical thinking. The attacker often instructs the victim not to call the family member directly (“my phone is broken” or “the police have my phone”) to prevent verification. They direct the victim to send money via wire transfer, gift cards, or cryptocurrency – methods that are difficult or impossible to reverse.

The CEO / Executive Impersonation

An attacker clones the voice of a company executive and calls an employee – typically in finance or accounting – to authorize a wire transfer or change payment details. This is the voice equivalent of business email compromise. The employee hears their boss’s voice giving a direct instruction and complies.

In one widely reported case, an energy company’s UK subsidiary wired $243,000 to a fraudulent account after the subsidiary CEO received a phone call from what he believed was his parent company’s CEO. The voice was a deepfake.

The Tech Support Call

Combining voice cloning with tech support scam tactics. The attacker clones the voice of someone from the victim’s IT department or a known tech support provider and uses it to gain remote access to the victim’s computer or extract credentials.

The Romance Scam Voice Verification

In long-running romance scams, victims sometimes request a phone call to verify the identity of the person they have been messaging. Voice cloning allows the scammer to produce a voice that matches the persona they have created, passing the “voice verification” test.

PanicVault app icon
Put this guide into practice PanicVault keeps your passwords in an encrypted KeePass file you own — offline-first, no subscription, on iPhone, iPad & Mac.
Get the app

Who Is Most Vulnerable

Voice cloning scams are effective against everyone, but certain groups are particularly targeted:

  • Elderly individuals – More likely to receive and trust phone calls, less likely to be aware of voice cloning technology, and more susceptible to emergency scenarios involving grandchildren.
  • Parents of young adults – The fear of a child in danger is one of the most powerful emotional triggers available to scammers.
  • Business executives and finance personnel – High-value targets where a single successful call can result in six-figure losses.
  • People with significant social media presence – More audio samples available for voice cloning.

How to Protect Yourself and Your Family

Establish a Family Code Word

The simplest and most effective defense against voice cloning scams is a family code word – a pre-agreed word or phrase that any family member can use to prove who they are. The FBI’s own advice is to agree on a secret word or phrase with your family. When you get an unexpected call from a “family member” in distress, ask for the code word. If they cannot give it, hang up and call them directly on a number from your contacts.

What makes a good code word:

  • Uncommon – Not something that would come up in normal conversation. “Pineapple” beats “hello” but is still too ordinary.
  • Memorable – Children and elderly relatives need to recall it under stress.
  • Not guessable from public information – No pet names, street names, schools, birthdays, or anything that appears on social media.
  • Not a security answer – Mother’s maiden name, first car and high school mascot are used for account recovery and may already be in data breaches.

Two unrelated words (“marble umbrella”) or a short phrase tied to a private family memory work well. Do not use any example you read in an article, including this one.

Setting it up:

  1. Choose it together, ideally in person. People remember what they helped create.
  2. Agree on the rule: any unexpected call about an emergency or money gets the code-word question. An excuse for not knowing it (“I forgot,” “just trust me”) means you hang up – a real family member will remember, because you practiced.
  3. Practice it. Call each other and ask for it, so asking feels normal rather than rude.
  4. Include everyone who might be impersonated or called: grandparents (the classic target of “grandchild in trouble” calls), college-age children, partners, and caregivers or babysitters.
  5. Refresh it every 6 to 12 months, and use the change as a reminder of the rule.

For elderly relatives, lead with a story rather than the technology (“scammers can make a call that sounds exactly like me”), frame it as protection rather than fear, and keep it to one word and one rule. If they might forget it, write it on an unlabeled card kept somewhere private.

Some families add a second, duress word that means “I’m saying I’m fine, but I’m not.” The same idea works outside the family too: agree on a verification phrase with business partners for unusual payment requests.

Verify Through a Separate Channel

If you receive a suspicious call – whether from a “family member,” a “bank representative,” or a “colleague” – hang up and call back on a known number. Do not use any number provided during the suspicious call. Look up the person’s or organization’s number independently.

For businesses, establish verification procedures that require out-of-band confirmation for financial transactions. A call from the CEO requesting a wire transfer should be verified by calling the CEO back on their known phone number or confirming in person. See how to verify a suspicious message for a complete verification framework.

Limit Your Audio Footprint

Consider reducing the amount of clear voice audio you post publicly:

  • Review social media posts and remove or limit access to videos where you speak clearly.
  • Set your voicemail greeting to a generic message rather than a personalized one.
  • Be aware that public speaking engagements, podcasts, and video interviews create material for voice cloning.

This does not mean you need to disappear from the internet. But awareness of what audio exists publicly helps you assess your risk level.

Recognize the Warning Signs

Even the best voice clones are deployed in scenarios with recognizable patterns:

  • Extreme urgency – The caller insists you must act immediately.
  • Secrecy – “Don’t tell anyone else” or “Don’t call them back” are red flags that the attacker is trying to prevent verification.
  • Unusual payment methods – Wire transfers, gift cards, cryptocurrency, or cash. Legitimate emergencies do not require payment via Apple gift cards.
  • Emotional manipulation – Crying, panic, fear. These are designed to override your rational thinking.
  • Inconsistent details – Ask specific questions that only the real person would know. The voice may be cloned, but the attacker does not have the person’s memories.

Use Technical Defenses

  • Two-factor authentication on all accounts protects you even if someone uses a cloned voice to talk you through giving up a password.
  • A password manager ensures your credentials are not vulnerable to voice-based social engineering. Your PanicVault vault unlocks with your master password or your own Face ID or Touch ID, so a caller cannot get into it, and its autofill will not hand your password to a fake site they send you to.
  • Call screening on your phone can help filter out unknown numbers. Both iOS and Android offer options to silence unknown callers, sending them to voicemail where you can assess the message without the pressure of a live call.

What to Do If You Receive a Suspected Voice Clone Call

  1. Stay calm. The emotional response is what the attacker is counting on. Take a breath.
  2. Ask for the family code word if the caller claims to be a family member.
  3. Ask specific questions that only the real person would know – details about recent events, shared memories, or inside jokes.
  4. Hang up and call back using the person’s real number from your contacts.
  5. Do not send money through any method until you have verified the situation through a separate channel.
  6. Report the attempt to local law enforcement and the FTC at ReportFraud.ftc.gov. See how to report phishing for detailed instructions.
  7. Warn the rest of your family that someone tried it – they may be next – and look at what public audio of the impersonated person the scammer could have used.

What to Do If You Have Already Paid

If you sent money before realizing it was a scam:

  • Contact your bank or payment provider immediately – Some transfers can be reversed if reported quickly.
  • File a report with the FTC at ReportFraud.ftc.gov.
  • File a report with the FBI’s IC3 at ic3.gov.
  • Contact local law enforcement – While recovery is difficult, reports help build cases and track patterns.
  • Do not blame yourself – Voice cloning scams are designed to be convincing, and falling for one is not a sign of naivety. Professional security experts have been fooled by these attacks.

The Technology Will Keep Improving

Voice cloning technology is on a trajectory of continuous improvement. The quality will get better. The required input audio will get shorter. Real-time voice conversion – where an attacker speaks normally and the AI converts their voice to the target’s voice in real time, allowing natural conversation – is already available and will become more accessible.

The defenses are human, not technological. Family code words, verification habits, and a willingness to pause and verify before acting under pressure. These defenses are simple, free, and effective regardless of how good the technology becomes. Set up your family code word today – it takes five minutes and could save you from devastating financial and emotional harm.

Protect Your Passwords with PanicVault

A secure, offline-first password manager using the open KeePass format. Your passwords, your file, your control.

Download on the App Store