The phone rings late at night. On the line is your son's voice — unmistakably his — crying, panicked. "Papa, I've had an accident, the police are involved, please send money now, don't tell Mummy." The voice breaks. A second man takes over, calm and official, and gives you a number to transfer to. Everything in you says act, protect your child, now.

Except your son is asleep in his hostel, unharmed. The voice you heard was cloned by a computer from a few seconds of audio lifted off a social-media video. The panic was manufactured to switch off the one thing that would save you: the pause to think.

This is the new face of fraud in India. The scams themselves are old — urgency, impersonation, a demand for money. What is new is that artificial intelligence now makes the lie sound and look real. The good news, and the point of this guide, is that the defences are old too, and they still work — if your family knows them before the call comes.

What AI Has Changed — and What It Hasn't

For years we told people: "If it's really your relative, you'll recognise the voice." AI has quietly broken that rule. But it has not broken the deeper logic of how scams work. Understanding both halves is the whole defence.

What AI has changed:


  • A voice can be cloned from a few seconds of a person speaking — a reel, a wedding video, a voice note. The clone can then say anything, in that person's voice, with emotion.

  • A face can be faked in video — "deepfakes" put a trusted person (a family member, a celebrity, even a news anchor or a government official) on screen saying words they never spoke.

  • Scam messages are now fluent and personalised — no more broken grammar. AI writes convincing text in Hindi, English, and regional languages, tailored to you.

What AI has not changed:


  • Every scam still needs you to act fast, in secret, and send money or share a code. That combination — urgency + secrecy + payment — is the signature of fraud, no matter how real the voice sounds.

  • No genuine emergency, bank, or authority is destroyed by a five-minute pause to verify. Only a scam is.

1930
National cyber-fraud helpline — call within the first hour
Source: I4C, Ministry of Home Affairs
Seconds
Of audio now enough to clone a person's voice
Source: Cybersecurity advisories, I4C
0
Legitimate emergencies that forbid you from calling back to verify
Source: Vidur Foundation
1
Family code word that defeats voice-clone scams
Source: Vidur Foundation

The New Scams, and How to See Through Them

The cloned-relative emergency call

A voice exactly like your child, grandchild, or sibling calls in distress — an accident, an arrest, a hospital, a kidnapping — and demands money immediately, in secret.

Defence: Hang up and call the real person back on their known number. If they don't answer, call someone else who can reach them. A cloned voice cannot survive a callback to the actual person. And agree a family code word in advance (see below) — a real relative in trouble can say it; a clone cannot.

Digital arrest, now with a fake uniform on video

The caller claims to be police, CBI, or customs — and now switches on video, showing what looks like a real officer in a real police station (a deepfaked or staged backdrop). They say a parcel in your name holds drugs, or your Aadhaar is tied to money laundering, and keep you on the call for hours until you transfer money to "clear your name."

Defence: There is no such thing as "digital arrest" in Indian law. No police force or agency arrests, interrogates, or settles cases over a video call, and none takes payment to a private account to "clear" you. However official the uniform looks on screen, disconnect and dial 1930.

The deepfake investment endorsement

A slick video shows a famous businessman, a film star, or a TV anchor "revealing" a scheme that "doubles your money" or a special app that "the government doesn't want you to know about." The face and voice are real-looking; the endorsement is entirely fake.

Defence: No real person guarantees you profits in a video ad. Genuine investments carry risk and never arrive through a forwarded video or a WhatsApp group. If returns are "guaranteed" and you're urged to act fast, it is a scam — regardless of whose face is on it.

The fake official / relative on WhatsApp

A WhatsApp account with your boss's, relative's, or an official's photo (lifted from social media) messages you: "I'm in a meeting, urgently buy these gift cards / transfer this amount, I'll return it." The picture is real; the account is not theirs.

Defence: A profile photo is not identity. Call the person's known number to confirm any money request. Real people do not conduct genuine emergencies through a brand-new WhatsApp number.

The Family Code Word: Your Strongest Single Defence

Here is a defence that costs nothing, needs no technology, and beats voice-cloning completely.

Agree a secret word or question with your family — something only you would know, never posted online. "What's the name of our first dog?" "Say the family code word." When any call claims to be a relative in trouble and asks for money, you ask for the code. A real relative knows it. An AI clone, however perfect the voice, does not.

Teach it especially to the elders in your family, who are targeted most and are most shaken by a crying voice. A grandmother who knows to ask "what's our code word?" before sending a rupee is safer than any software could make her.

The rule that survives every new technology
Pause. Verify on a channel you trust. Then act.
Every AI scam depends on you acting inside the first few panicked minutes, without checking. So build one unbreakable habit: never act on an urgent money demand from an incoming call or message. Hang up. Call the person or institution back on a number you already have — from your contacts, the back of your card, or the official website. The scam cannot follow you onto a channel you chose. This single habit defeats voice clones, deepfakes, and every scam not yet invented.

When You Suspect a Deepfake — Small Tells

AI is good, but not perfect. In a suspicious video call, look for:


  • Unnatural blinking, stiff mouth movement, or lips slightly out of sync with the words.

  • Odd lighting or blurring where the face meets the hair or neck.

  • A refusal to do simple live actions — ask the person to turn their head sideways, wave a hand in front of their face, or say a random sentence you choose. Real-time deepfakes often break under these.

But do not rely on spotting the fake. The tells are getting fainter every year. Verification by callback and code word is the defence that lasts, because it doesn't depend on the fake being imperfect.

If Money Is Already Gone — The Golden Hour

Speed decides whether stolen money can be frozen. The moment you realise you've been defrauded:

  1. Call 1930 — the national cyber-fraud helpline. Reported fast, the money can often be frozen as it moves through the fraudsters' accounts.
  2. File at cybercrime.gov.in with screenshots, numbers, and transaction IDs. Save the acknowledgement number.
  3. Call your bank on its official number and freeze the account/card/UPI used.
  4. Keep everything — call logs, recordings, chats, transaction details — as evidence.

Report even if you feel embarrassed, and even for a small amount. Your report helps shut down the accounts before they take someone else's savings. (For the full fraud-response playbook and your rights on refunds, see our UPI fraud protection guide.)

Two things to do at dinner tonight

One: agree a family code word with everyone — parents, grandparents, children studying away from home — and make clear that no money moves in an "emergency" until someone says it. Two: save 1930 in every family phone as "Cyber Fraud Helpline," and teach the elders the single rule: a real emergency can always survive you hanging up and calling back. Fifteen minutes at the dinner table is worth more protection than any app.

What You Can Do

  • Set a family code word today — the one defence voice-cloning cannot beat.
  • Teach the elders the callback rule: never send money on an incoming call; hang up and call back on a known number.
  • Treat every "guaranteed returns" video as a scam, whoever's face is on it.
  • Lock down what scammers harvest: keep social-media profiles and voice/video posts private where you can — that's the raw material for clones.
  • Save 1930 in every family phone, and report fast if money is lost.
  • Tell one elder in your family how voice cloning works. The shock of learning it now, calmly, is what stops the panic later.

Sources