AI voice cloning and deepfake scams: spot and report

What the scams look like

A deepfake is a face, video or voice made or altered with AI to look and sound real. Cloning only a voice is sometimes called a voice clone. The FTC says a scammer can clone a voice from a short audio clip, for example one taken from a video posted online, so ordinary people can be targeted.

• Family emergency call

  How it plays out

  A voice like your child's or grandchild's says they're in an accident or in jail and need money now

  Warning signs

  Unfamiliar number, panic and pressure, “don't tell anyone,” payment by wire, crypto or gift cards

• Fake kidnapping

  How it plays out

  You hear a cloned voice or see a synthetic video and a ransom is demanded

  Warning signs

  Told not to contact police, demand for immediate payment

• Video call from someone you know

  How it plays out

  A friend or boss appears on video and asks for money or gift cards

  Warning signs

  Poor quality, short or choppy call, odd speech, lips out of sync with audio

• Celebrity or expert investment pitch

  How it plays out

  A famous investor's face and voice promote a “guaranteed” opportunity

  Warning signs

  Promises of high returns, an invite to a private group chat, a request to install an unfamiliar app

What the FTC and FBI say

The FTC's consumer alert on AI-enhanced family emergency scams says to treat a familiar voice as unreliable. Its advice is to hang up and call the person back on a phone number you know is theirs, and to check with another relative or friend if you can't reach them. It also lists wire transfers, cryptocurrency and gift cards as payment requests that should raise a red flag. On its Family Emergency Scams page, the FTC notes that scammers may pose as a lawyer, police officer or doctor to seem credible.

The FBI's December 2024 public service announcement says criminals use generative AI to make fraud more convincing. Its tips include agreeing on a secret word or phrase with your family, keeping social media accounts private, hanging up and verifying through a number you find yourself, and not sending money, gift cards or cryptocurrency to people you know only online or by phone.

The first move: hang up and call back

Changing how you verify works better than trying to judge whether a voice is real. Cloned voices keep getting better, so your ear is not a dependable test. Practice this sequence.

• If the caller asks for money, gift cards, account numbers, verification codes or an app install, hang up.
• Call back using the number saved in your own phone, not the one the caller gave you. If they don't answer, try another family member, their friends, their school or their workplace.
• “I can't talk right now,” “don't tell anyone” and “send it immediately” are classic scam lines.
• If someone claims to be a bank or agency, look up the official number yourself and call that.
• Caller ID can be faked, so a familiar name or number on screen doesn't prove who is calling.

Set a family safe word

People panic on these calls, so the plan has to exist before the call comes. Spend five minutes setting it up when the family is together. The FBI recommends this approach in its public service announcement on AI-enabled fraud.

• Pick a word or phrase that isn't on social media or in photos and that only your family knows. For example, “the name of our first cat.”
• Share it in person or by voice, not in a family group chat. If an account is hacked, the word is exposed too.
• Change it now and then, and right away if you think it leaked.
• Teach parents and kids the rule: if the caller can't give the safe word, you hang up.

Hi Mom, if you ever get a call or video that sounds like me asking for money or gift cards, hang up and call me back on my number. If I don't answer, check with Jordan first. If the caller can't give our safe word, it isn't me.

What to look for on a video call

Treat visual clues as a backup only. As the technology improves, these signs may disappear.

• The edges of the face or hair shimmer, blur or flicker.
• Mouth movements are slightly out of sync with the voice.
• Blinking looks odd, or the eyes stay fixed in one direction.
• Asking the person to turn their head or put a hand in front of their face can make the image glitch.
• The call is short and choppy, with an excuse about a bad connection.

Investment pitches and celebrity impersonation

• If a famous person in an ad says “make thousands a month with this method,” assume it's an impersonation.
• “Guaranteed returns,” “risk-free” and “insider tips” are warning words.
• If someone says they work for a brokerage, check with that firm directly using contact details from its official website.
• Be wary of anyone who asks you to install an unfamiliar trading app or move to a private group chat.
• Screenshot the chat, messages and account details before you delete anything. You'll need them for a report.

If you were targeted or lost money

• United States: report scams to the FTC at reportfraud.ftc.gov. For internet-enabled crime, including AI-assisted fraud, file a complaint with the FBI's Internet Crime Complaint Center at ic3.gov. The FBI asks you to include details such as the scammer's name, phone number, email, payment type, amounts, account numbers and how the contact started.
• United Kingdom: Report Fraud replaced Action Fraud in December 2025 as the national reporting service for England, Wales and Northern Ireland. Its website is reportfraud.police.uk and its phone line is 0300 123 2040. Scotland reports to Police Scotland on 101 for now.
• Elsewhere: use your country's consumer protection agency or police. Search for the official government site and avoid links from messages or ads.

Sexual deepfakes and non-consensual images

Making fake sexual images of a real person without consent harms them, and laws on this differ by country and state and are changing. If you are a victim, you don't have to deal with it alone. Save the links, account names and screenshots before reporting, but don't open or forward the material more than you need to.

StopNCII.org is a free tool run by the Revenge Porn Helpline, part of the UK charity SWGfL. It creates a digital fingerprint (a hash) of an image on your own device, so the image itself isn't uploaded. Participating companies can then use the hash to detect and remove matches. To use it, you must be the person in the image, be over 18 now, and have been 18 or older when it was taken. The site doesn't say whether fully AI-generated images are covered, so check its guidance.

Take It Down, from the National Center for Missing & Exploited Children (NCMEC), is a free service for nude or sexually explicit images taken of a person under 18. It works the same way, with a hash made on your device. It only works on participating platforms, and NCMEC suggests also reporting to its CyberTipline for further help. The site doesn't say whether AI-generated images are covered. If the images were taken when you were 18 or older, it points you to StopNCII.org.

Reduce your exposure

• Limit who can see videos with your voice, especially longer clips. The FBI advises keeping social media private and limiting followers to people you know.
• Check who is calling before you accept a video call from an unknown number.
• Post fewer public photos of your children.
• Don't open investment ad links from unknown sources.
• Share this article with your parents and set a safe word together.

Related terms

Deepfake

A face or video made or altered with AI to look real.

Voice clone

An AI copy of a specific person's voice.

Safe word

A secret word or phrase a family agrees on to check who is really calling.

Hash

A digital fingerprint of a file. Services like StopNCII.org and Take It Down share the hash, not the image.

FAQ