Deepfakes and AI Scams: How to Recognize Them and Protect Your Family

The voice on the phone was your child's, crying, asking for money to solve an emergency. The video was a famous TV host recommending an "approved" investment platform. The voice message was your boss ordering an urgent wire to a supplier. None of the three was real, and all three scams happen every day.

The technology to clone voices and faces became good, cheap and accessible: a few seconds of public audio (a story, a birthday video) are enough to generate a convincing voice. And in an election year, this machine works double shifts. The good news: the scams follow known patterns, and you can defend yourself without becoming paranoid.

Scam 1: the relative in trouble

How it works: a call or voice message with the cloned voice of a child or sibling, always in an emergency: accident, kidnapping, stolen phone. Emotion switches off critical thinking, and the request is always fast money, now.

What gives it away: extreme urgency, secrecy ("don't tell anyone"), an unknown number ("my phone broke") and refusal to answer questions outside the script.

Scam 2: the celebrity endorsing an investment

How it works: videos with the face and voice of TV hosts, journalists and business figures promoting investment platforms or giveaways, often as paid ads on social networks, which lends fake legitimacy.

What gives it away: guaranteed returns (they do not exist), rush ("last spots"), a site outside official channels, and the question itself: why would a billionaire advertise your path to wealth?

Scam 3: the boss ordering a payment

How it works: an audio or call with the voice of the owner or CFO orders an employee to make an urgent, confidential transfer. In companies, this is the scam that moves the most money per hit.

What gives it away: a break from the normal payment process, demanded secrecy, hierarchy pressure and an unusual channel (the boss who always emails suddenly sends voice notes).

The defenses that actually work

• A family code word. Agree on a word or question only your family knows. Zero cost, kills the cloned-voice scam instantly. Do it today; it takes two minutes in the family group chat.
• Hang up and call back. Strange request from any number? Hang up and call the number YOU have saved. The scammer answers on their number; the real person answers on theirs.
• Urgency + secrecy = alarm. "It has to be now" combined with "don't tell anyone" is the signature of almost every scam, AI or not. Treat it as a siren.
• Second channel, always. Money or sensitive-data requests get confirmed on ANOTHER channel: audio gets confirmed by phone call, chat gets confirmed in person. In companies, make it written policy: payments only with confirmation through an independent channel, no exceptions even for "the board".

How to check a suspicious video

• Look at edges and hands: deepfakes still stumble on hairlines, teeth, hands and fine lip-sync.
• Find the official source: if the celebrity really launched that, it is on their verified channels and in the press. If it only exists in the ad, it is a scam.
• Question the context: who gains from you believing it? A bombshell video with zero serious news coverage is a red flag.

And an honest rule: visual detection gets harder every month. The durable defense is not "watching the video more closely"; it is verifying outside the video, at the source.

Election years: double the attention

A candidate's "leaked" audio, a "fraud" video, a "censored" poll: the electoral deepfake menu is predictable. Before sharing any bombshell, take three seconds: did any professional outlet report it? Electoral authorities, fact-checking agencies and official channels exist for this. Sharing a lie, even in good faith, is working for free for the scammer.

If you (or someone) fell for it

Act fast: contact the bank immediately to try to reverse the transfer, file a police report online, keep screenshots and numbers, and warn the family group or the company to stop the wave (scammers repeat targets). Shame does not recover money; speed does.

At bottom, the tool changed but the scam is the same as ever: hijacking your emotion to outrun your verification. The defense is also the same as ever, with one new layer: trust people, distrust channels, and set up that code word today.