Free public service. Run by Apex Influence, a private marketing company. Not the police or a bank. The buttons on these pages go to official government channels. We will never ask you for money or an OTP.
Scam Help / AI voice clone and deepfake scams: how to check a call before you pay

Scam Help · Free public service

AI voice clone and deepfake scams: how to check a call before you pay

Yes. Scammers can copy a voice from a short clip and call you as your son, niece or boss. Hang up. Call the person back on the number you already have. Ask for your family safe word. Never send money while the caller is still on the line. Your real relative will still be there when you call back.

We are a private marketing company. We are not the police or a bank. We cannot recover money for you. This page explains how the trick works and what official agencies advise.

How does an AI voice clone scam work?

The scammer needs a short sample of a real voice. It can come from a video, a voice note or a social media post. A cloning tool turns that sample into a voice that sounds like your relative. The FBI says criminals generate short audio clips of a loved one's voice and use them to pose as a relative in a crisis.

  1. You get an unexpected call, voice note or video call. Often it is from a new number.
  2. The caller says their own phone is broken or lost. That is why the number is new.
  3. A crisis story follows: an accident, an arrest, a hospital bill, or a kidnapping.
  4. Pressure is added: act now, tell no one, do not call anyone else. This is meant to stop you from checking.
  5. Payment is demanded in a way that is hard to reverse: bank transfer, crypto or gift cards.
  6. Sometimes a second, bigger request comes after the first payment.

Source: FBI Internet Crime Complaint Center advisory PSA241203, 3 December 2024; US Federal Trade Commission consumer alert, March 2023.

Is this scam really happening? What do the numbers say?

Yes. Here is what official agencies reported. Read each number as "reported by" that agency, for that period.

What was countedReported figureWho reported it
Complaints with AI-related informationMore than 22,000 complaints (the report's chart shows 22,364)FBI IC3, calendar year 2025
Adjusted losses on those complaints$893,346,472FBI IC3, calendar year 2025
Losses to "distress" scams, where a loved one's voice is imitatedOver $5 millionFBI IC3, calendar year 2025
All fraud reports to the US FTC3 million reports, $15.9 billion reported lossesUS FTC, calendar year 2025
Imposter scams to the US FTCMore than 1 million reports, more than $3.5 billion reported lossesUS FTC, calendar year 2025

Source: FBI IC3 2025 Internet Crime Report, page 39; US FTC statement to the Joint Economic Committee, March 2026. Calendar year 2025.

The FBI also says the distress scam is widening. It no longer targets only grandparents. It now copies other family members and close friends in different kinds of emergency.

Careful: the FBI and FTC numbers come from different systems and overlap. Do not add them together. We found no separate count of voice-clone scams from India's government. The national figures below cover all cyber fraud.

What does a real case look like?

Case 1: "I met with an accident", Lucknow, India, 2023

What happened. A man in Lucknow got a call in what sounded like his brother-in-law's voice. The caller said he and a friend had met with an accident. The friend was badly injured. His own phone was damaged, so he was using someone else's. The man panicked and transferred money for the treatment. Police said the voice was cloned using AI. ETV Bharat, citing police, reported the loss as over Rs 1.5 lakh.

The tell. A new number, a reason why the real phone was not working, and a quick demand for money.

What would have helped. He later called his real relative and learnt it was fake. Making that call before paying would have shown the call was false. He filed an FIR. The investigation was continuing at the time of the report.

Source: ETV Bharat, December 2023, reporting police.

Case 2: a second Lucknow report, 2023

What happened. A 25-year-old in Lucknow received a call from someone who passed himself off as a relative. The distress story was told in a voice that matched. He transferred Rs 44,500 to the caller's account, as reported by IANS via Gulf News. IANS described it as the first AI-generated voice fraud case reported in Lucknow.

The tell. A "distant relative" in trouble, asking for a bank transfer.

What would have helped. A call-back to the relative on a saved number, or a safe word.

Source: Gulf News (IANS), December 2023.

Case 3: fake bosses on a video call, Hong Kong, 2024

What happened. A finance employee at a multinational firm joined a video conference. The people shown as senior officers were fake. The employee was asked to move money to named bank accounts. Hong Kong police said about HK$200 million (US$26 million per AFP) was lost through 15 transfers. Police got the report on 29 January 2024. No arrest had been made at the time.

The tell. A request for many transfers to named accounts, with the pressure of a "boss" on screen.

What would have helped. Checking each transfer with the real officer through a separate, known channel. Treat a face on a screen like a voice on a phone: it proves nothing alone.

Source: Hong Kong Free Press (AFP), citing Hong Kong police, published 5 February 2024.

Case 4: a fake famous face selling an investment, Bengaluru, 2024 to 2025

What happened. A chartered accountant in Bengaluru saw deepfake videos on social media. Well-known business figures appeared to promote a stock investment. He paid in. Later, withdrawals were denied or met with demands for more money called taxes and charges. Deccan Herald reported the loss as Rs 23.20 lakh. He first saw the videos on 2 October 2024. Police registered the case on 13 January 2025 and said they were tracing the money through the banks involved.

The tell. A famous person promoting an investment, then a second request for money to take your own money out.

What would have helped. Stopping at the first demand for "tax" before any withdrawal. A demand for more money to release your own money is a warning sign.

Source: Deccan Herald, reporting police, 2025.

What are the warning signs?

Source: CERT-In Advisory CIAD-2024-0060, 27 November 2024; FBI IC3 PSA251205, 5 December 2025; US FTC consumer alert, March 2023.

What should I do in the next hour?

Use this checklist. Decision points are in bold.

  1. Stop. Do not pay yet. The FBI says scammers push a false sense of urgency. A short pause is your best tool.
  2. Hang up. Do not press any link. Do not stay on the line.
  3. Call the person back on a number you already know is theirs. The FTC says do not trust the voice.
  4. Did they pick up and say all is fine? It was a scam. Go to the reporting step.
  5. No answer? Try another relative or their friends. The FTC advises this too.
  6. Ask the caller your safe word, if your family has one. A cloned voice should not know it.
  7. Is it a kidnapping story? Try to reach your loved one first, before any payment. Take screenshots of any photo or video sent. The FBI advises this.
  8. Already paid? In India, call 1930 right away. Speed matters, see the next section.

Source: FBI IC3 PSA251205, 5 December 2025; US FTC consumer alert, March 2023; PIB (Ministry of Home Affairs) release, July 2026.

I already sent money. How fast must I report in India?

As fast as you can. In India, report online financial fraud on the National Cyber Crime Reporting Portal (cybercrime.gov.in) or call helpline 1930. Do not wait for the money to come back on its own, and do not wait to be sure.

Be honest with yourself about the odds. Here are the official numbers. They cover all cyber fraud, not only voice clones.

Source: PIB (Ministry of Home Affairs), National Cybercrime Response Mechanism release of 22 July 2026; PIB release of 28 July 2026.

Reporting early is still the right step. It is not a promise. Nobody can promise you your money back, and we cannot do it either. For the exact steps, see what to do in the first hour.

How do I set up a family safe word?

The FBI advises a secret word or phrase that your family agrees on to check each other's identity. Do it this week.

  1. Sit together in person or on a call you trust. Do not agree it by text or on social media.
  2. Pick a word or phrase nobody could guess. Do not use a pet's name, a birthday or a street name that is public online.
  3. Tell every family member, including grandparents and teenagers.
  4. Agree the rule: if anyone asks for money in a hurry, ask for the safe word first.
  5. Agree a second rule: if the word is not known or the caller refuses, hang up and call back on a saved number.
  6. Practise once. Call each other and ask for the word, so nobody feels shy on a real day.

Source: FBI IC3 PSA241203, 3 December 2024; FBI IC3 PSA251205, 5 December 2025 (code word for virtual kidnapping).

What should I do over the next week?

Source: CERT-In Advisory CIAD-2024-0060, Deepfakes: Threats and Countermeasures, 27 November 2024 (rated High severity).

What is India doing about deepfakes?

On 10 February 2026 the IT Rules were amended for synthetically generated information, including deepfakes. The change calls for clear labelling and traceable metadata for permitted AI content. It also cuts the time to remove such content from 36 hours to 3 hours when a valid government or court notice is given. Helpline 1930 and cybercrime.gov.in remain the reporting routes.

Source: PIB (Ministry of Electronics and IT) release of 6 August 2026, amendment dated 10 February 2026.

Law enforcement is also acting across borders. In December 2025 the CBI, acting on FBI inputs, arrested six people at an illegal call centre near Delhi that posed as US agencies. The CBI put the amount defrauded at USD 8.5 million, and Rs 1.88 crore in cash was recovered. This was a government impersonation scheme, not a confirmed voice-clone case. We mention it because it shows India's own agencies are chasing these networks.

Source: The Tribune, reporting the CBI press release of 12 December 2025.

What to do now

Where can I read more?

Common questions

Can scammers really copy my voice or my relative's voice?

Yes. The FBI says criminals generate short audio clips of a loved one's voice and use them to pose as a relative in a crisis. The clip can come from content posted online. So treat any urgent call asking for money as unproven until you check it.

What should I do if I get a call from a family member asking for money?

Hang up. Call the person back on a number you already know is theirs. If you cannot reach them, try another relative or their friends. Ask for your family safe word. Do not send money until you have spoken to the real person.

What is a family safe word and how do I choose one?

It is a secret word or phrase your family agrees on to check each other's identity, as the FBI advises. Agree it in person or on a trusted call, not by text. Make it unguessable, tell everyone, and ask for it whenever someone requests money in a hurry.

How can I spot a deepfake video?

CERT-In lists lip movements and sound out of step, odd blinking or expressions, uneven lighting, robotic-sounding voices and distorted visuals. Also verify the source and cross-check with other reliable sources before you act.

I already paid after a fake family emergency call. What now?

In India, call helpline 1930 and report on cybercrime.gov.in right away, because speed matters. Refunds are not assured: the Ministry of Home Affairs reported more than Rs 206 crore refunded against more than Rs 56,087 crore reported, over financial years 2023-24 to 2025-26. We cannot recover money.

Is the Indian government counting voice-clone scams separately?

We found no separate count. The Ministry of Home Affairs figures cover all cyber fraud, so they should not be read as AI-scam numbers. The FBI did report more than 22,000 AI-related complaints in the United States in 2025.

Sources checked 3 October 2026

Figures are as reported by the named agencies for the periods shown. Reporting rules differ by country, so totals are not directly comparable.