A phone call from a panicked relative, a video message from your manager, a text about a parcel that cannot be delivered: in the run-up to Christmas, scams arrive in every form, and AI has made them far more convincing. Cloned voices and deepfake video are no longer research curiosities. This guide explains how these scams work, the warning signs that still give them away, and the simple habits that protect you and your family.
How voice cloning works
Voice cloning uses a model trained on thousands of hours of human speech. Given a short recording of one person, it learns the characteristics of that voice, such as pitch, accent and rhythm, and can then read out any text in something close to it. Some tools work almost in real time, so a scammer can type replies during a live call.
The recording does not have to be long. In a 2024 warning about voice cloning scams, Starling Bank said that three seconds of audio can be enough to produce a usable clone. Voice notes, videos on social media and even a recorded voicemail greeting are all possible sources. These are the same generative AI tools that draft emails and summarise documents in offices every day, pointed at a very different goal.
The scams to expect this winter
The technology changes, but the stories scammers tell are familiar. In a public service announcement published in December 2024, the FBI described how criminals use generative AI to make old frauds more believable. The patterns that matter most for ordinary households are these:
- The family emergency. A call or voice note in a loved one's voice says they have been in an accident, arrested or robbed abroad and need money now. Holiday travel makes this story easy to believe.
- The boss or colleague. A voice message or video call appears to come from a senior manager asking for an urgent payment or a change to a supplier's bank details, often just before a holiday when people are rushing to clear their desks.
- The bank or the police. A caller who sounds official tells you your account is at risk and asks you to move money to a "safe" account.
- The celebrity endorsement. Deepfake videos of famous people promoting investment schemes, giveaways or heavily discounted products, timed for Black Friday and Christmas shopping.
- The fake charity. AI-generated images of disasters used to collect donations for appeals that do not exist.
Deepfake video is no longer limited to pre-recorded clips. In 2024 the engineering firm Arup confirmed that an employee in its Hong Kong office had transferred around HK$200 million after a video conference in which the other participants, including someone appearing to be a senior executive, were fakes. If that can happen in a large company, a convincing call to a grandparent is well within reach.
Warning signs that still give fakes away
Detection is getting harder, and you should not rely on spotting technical flaws. Still, the following signs are worth knowing:
- Urgency and secrecy. "Don't tell Mum", "I need it in the next hour", "the bank must not know". Pressure is the scammer's main tool, with or without AI.
- An unusual way to pay. Requests for bank transfers to a new account, gift cards or cryptocurrency are a red flag whatever the voice sounds like.
- Odd phrasing or flat emotion. The FBI advises listening closely to tone and word choice. A cloned voice may sound right but use expressions the real person never would, or stay strangely calm in a supposed crisis.
- Visual glitches in video. Distorted hands, unnatural eyes or teeth, inconsistent shadows and lips slightly out of sync with the words.
- A reason not to call back. A broken phone, a borrowed handset, a meeting that cannot be interrupted. Anything that stops you checking is suspicious.
Five habits that protect your family
- Agree a safe phrase. Choose a word or short phrase with close family that anyone can ask for on an unexpected call. Agree it face to face and never write it in a message or post. Both the FBI and Starling Bank recommend this, and it works against even a perfect voice clone.
- Hang up and call back. Use a number you already have saved, not one given during the call. For banks in the UK, dialling 159 connects you to many major banks through a number you know is genuine.
- Slow down on payments. Make it a household rule that no money moves on the strength of a single call or message. A genuine emergency will survive a five-minute check.
- Limit what is public. Consider making social media accounts private, especially those with videos of you or your children speaking, and review who can see older posts.
- Talk about it now. Older relatives are often the main target. A ten-minute conversation over the holidays about how these calls sound is one of the most effective protections there is.
At work: a checklist before you pay
Businesses are targeted by the same techniques, often with larger sums at stake. Before acting on any request to move money or change payment details, check:
- Has the request been confirmed through a second, separate channel, such as a call to a known number or a word in person?
- Does it follow the normal approval process, or does it ask you to skip it?
- Would the person asking normally contact you this way, at this time?
- Are you being asked to keep it confidential from colleagues who would usually be involved?
Clear rules help staff say no without feeling they are defying a senior colleague. Organisations that use AI responsibly also think about this side of the technology, a theme we explored in our article on responsible AI, transparency and accountability.
If you think you have been targeted
If you have sent money or shared bank details, contact your bank immediately using the number on the back of your card or 159. The sooner a payment is reported, the better the chance of stopping or recovering it. Report the scam to the police fraud reporting service in your country, and in the UK forward suspicious text messages to 7726, the free number that passes them to your mobile network. Warn the person who was impersonated as well, because their voice or image may be used again.
Frequently asked questions
Can a scammer clone my voice from a short phone call?
Potentially, yes. A few seconds of clear speech can be enough for current tools, which is one reason to be careful about long conversations with unknown callers who keep you talking.
Are there apps that detect deepfakes reliably?
Detection tools exist, but none is reliable enough to depend on for a phone call. Verification habits, such as a safe phrase and calling back on a known number, work regardless of how good the fake is.
Is it safe to post videos of my family?
Posting is a personal choice, but public videos give scammers material to work with. Sharing them with a private group of people you know reduces the risk considerably.
The most useful step you can take today is the simplest: pick a safe phrase with the people closest to you, and tell them why. It costs nothing, takes a minute over dinner, and turns the most convincing AI-generated call into one you can end with a single question.
