Scam Awareness
Deepfake Scams: How to Verify Voice and Video Impersonation
Learn how deepfake scams use cloned voices and synthetic video, how to verify urgent requests, and what Indian readers can do to preserve evidence and report suspected fraud.
Deepfake scams use AI-generated or manipulated audio, video, or text to make an impersonator appear credible. The safest response is not to win an argument about whether a voice or face is genuine. Pause, refuse an unverified transfer or disclosure, and confirm the request through a trusted channel you already know. The FBI has documented AI-generated audio used to imitate a loved one in a crisis, seek money or ransom, or reach bank accounts. It has also documented AI-generated video in real-time chats with alleged executives, law-enforcement officers, or other authorities. These examples describe scam methods, not proof of a particular level of prevalence in India. (FBI Internet Crime Complaint Center, 3 December 2024.) This guide explains how deepfake scams work, why visual or audio glitches are not a reliable verdict, and what to do when a request feels urgent. For India-focused context, our [India cybersecurity guide](https://cyberoinfo.com/cybersecurity/india-cybersecurity) covers the wider reporting and cyber-safety framework. The practical rule here is simple: trust an independently verified relationship or number, not the realism of the media alone. > **Key Takeaways > - A familiar voice or convincing video is not enough to authorise a payment, OTP, password, or identity disclosure. > - Verify through a pre-existing number or channel, and use a family or team secret phrase for urgent requests. > - Treat urgency, secrecy, a new platform, and pressure to bypass normal checks as behavioural warning signs. > - In India, preserve the message and transaction details, and report suspected financial fraud through 1930 or cybercrime.gov.in. > - Audio and video artefacts are clues at most; their absence does not prove that a contact is genuine.
How deepfake scams use cloned voices
A voice cloning scam usually starts with trust. The caller or message may sound like a relative, colleague, manager, recruiter, bank representative, or public official. The request then adds pressure: an emergency, a confidential transfer, a password reset, a ransom demand, or a move to a different messaging platform. The FBI says criminals have used AI-generated audio for these forms of impersonation and for attempts to access bank accounts. (FBI Internet Crime Complaint Center, 3 December 2024.)
The audio is often only one part of a larger social-engineering sequence. In a May 2025 public service announcement, the FBI described messages that used text and AI-generated voice while impersonating senior US officials. The campaign sought to build rapport, move targets elsewhere, and potentially obtain account access or information. That sequence matters because a believable voice can make a later link or request feel familiar, even when the conversation began from an unknown number. (FBI Internet Crime Complaint Center, 15 May 2025.)
Do not treat a caller ID, a known name, or a short personal detail as independent proof. If the request involves money, an OTP, a password, identity data, or a change to normal payment instructions, end the call and contact the person through a number already saved in your records. A separate call is stronger than asking the same caller to confirm their identity.
A pre-agreed family or team phrase can add another layer. The phrase should not be sent in the same conversation or stored where an attacker who has taken over an account can read it. It is a verification aid, not a guarantee: if the person is unable to use it, slow down and use another trusted route before acting.
How deepfake scams use synthetic video
A deepfake video scam can use a live-looking call, a private message, a recorded clip, or a video presented as proof that a contact is real. The FBI has described real-time chats involving alleged executives, law-enforcement officers, or other authorities, as well as private messages used to persuade targets that an online contact is genuine. It has also warned about misleading investment promotions using AI-generated video. (FBI Internet Crime Complaint Center, 3 December 2024.)
The video may be paired with phishing and a fake website. Microsoft’s April 2025 Cyber Signals report says attackers combine deepfakes and voice cloning with phishing and fake websites to appear legitimate at scale. In its discussion of AI-generated interviews, Microsoft mentions lip-sync delays, robotic speech, and unusual facial expressions as possible warning signs. These are observations to examine, not a consumer test that can prove a video is fake. (Microsoft Cyber Signals Issue 9, 16 April 2025.)
A smooth video is not proof of authenticity, and an awkward pause is not proof of manipulation. A poor connection, compression, lighting, disability, or ordinary camera behaviour can create similar artefacts. The more dependable question is whether the request fits an independently confirmed relationship and a normal process.
Ask for time rather than a performance. Do not ask the person on the suspicious call to provide more private information as a “test.” Instead, stop the conversation, call a known number, check with a colleague through the usual directory, or verify a business instruction using an established procedure. A face on a screen should never replace access controls or a second-person approval for a sensitive action.
When a voice or video sounds completely real
When media sounds real, focus on behaviour and context. Urgency, secrecy, a new number or platform, an unexpected request for money or credentials, and pressure to skip normal verification are meaningful risk signals when they appear together. None proves fraud alone, but each is a reason to pause before an irreversible action.
Use an independent channel that the caller did not provide. A family member can be reached on a saved number, a colleague can be checked through the company directory, and a payment instruction can be confirmed using the organisation’s normal approval route. Do not click an unverified link to perform that check. The FBI has specifically advised independent verification and avoiding unverified links in its guidance on impersonation campaigns. (FBI Internet Crime Complaint Center, 15 May 2025.)
Limit the information that can be collected about you in advance. The FBI recommends limiting public image and voice material and reducing unnecessary social-media exposure. This will not prevent all impersonation, but it can reduce the material an attacker may use to make a message feel personal. Review what is public without assuming that privacy settings can eliminate the risk.
If you are unsure, delay the transaction. A legitimate family member, employer, bank, or authority should have a safe way to tolerate a short verification step. A demand that you must act immediately, keep the request secret, or move to a new platform is a reason to stop and consult someone you trust. These safeguards are useful even when no synthetic media is involved.
What to do after a suspected deepfake scam
If money, credentials, or identity information may have been exposed, first stop further contact and do not send more funds. Contact the bank, payment provider, or relevant account owner through an official channel and ask what containment steps are available. Change exposed passwords from a clean device, enable MFA where possible, and review active sessions or recovery details. If the incident also involves a broader data exposure, our [data breach response guide](https://cyberoinfo.com/cybersecurity/data-breach-response) provides a separate first-hours framework.
Preserve evidence before deleting or editing it. Keep the original message, phone number or account identifier, timestamps, URLs, payment references, wallet or bank details, screenshots, and any available audio or video. Do not forward suspicious media widely. Store copies in a safe place and record what happened in time order; this helps you explain the sequence to a bank, platform, employer, or reporting authority.
In India, suspected financial cyber fraud can be reported through **1930** and **cybercrime.gov.in**, as identified in the I4C reporting ecosystem. The Ministry of Electronics & Information Technology, through PIB, describes deepfakes as potentially involving synthetic audio, video, and text and notes their possible effects on dignity, reputation, and privacy. Its release identifies CERT-In, I4C, the National Cyber Crime Reporting Portal, and 1930 within the response framework. (PIB/MeitY, 8 August 2025.)
Keep organisational duties separate from individual help. CERT-In directions and incident processes may apply to covered organisations, while an individual victim’s immediate priority is containment, evidence preservation, and reporting through the appropriate route. If an account or identity is at risk, use the service provider’s official recovery process as well. Our [identity attacks and MFA bypass guide](https://cyberoinfo.com/identity-security/identity-attacks-mfa-bypass) explains why changing a password alone may not close every active session or token.
How families and organisations can prevent deepfake scams
Prevention works best when verification is designed before an urgent message arrives. Families can agree on a secret phrase and a second contact route. Organisations can require callback verification, dual approval for unusual payments, and a known directory for executives, recruiters, suppliers, and support staff. These controls reduce the chance that a convincing voice or video becomes the only evidence for a sensitive decision.
Keep high-risk requests inside established workflows. A bank-detail change should use the normal vendor process; a senior-official request should be confirmed through a separate office channel; and a recruiter or interview contact should not require credentials or payment before identity is verified. Training should rehearse the pause-and-check sequence without shaming people who receive a realistic message.
Reduce unnecessary exposure of voice and image material, especially when public posts reveal names, relationships, workplaces, travel, or family routines. The goal is not to disappear from the internet or to blame people for having public profiles. It is to make sensitive context harder to assemble and to make a false sense of familiarity less persuasive.
Layer these measures with account security. Use MFA, protect recovery methods, keep devices and applications updated, and review unexpected sign-in or payment alerts. If an account is impersonated, report the content through the applicable platform channel while preserving the original evidence. No single setting or deepfake detector replaces independent verification.
FAQ: deepfake voice and video scams
### Can a strange voice or lip-sync delay prove a deepfake?
No. Robotic speech, lag, unusual expressions, or lip-sync problems may be clues, but they can also result from network conditions, compression, lighting, or ordinary camera behaviour. The FBI’s recommendations to scrutinise audio and visual irregularities should not be read as a guarantee that artefacts will reveal synthetic media. (FBI Internet Crime Complaint Center, 3 December 2024.)
### What should I do if a loved one asks for money in a cloned voice?
Pause and do not transfer money or share an OTP, password, or identity detail. End the call and contact the person on a number already saved or through another trusted family member. A secret phrase agreed in advance can help, but independent callback verification remains important.
### Should I trust a video call from a bank, police officer, executive, or government official?
Do not authorise a sensitive action from the video alone. Stop and verify the person and request through the organisation’s known public or internal channel. A legitimate process should not require bypassing normal identity, payment, or approval checks because a video call appears convincing.
### Where can I report a deepfake scam in India?
For suspected financial cyber fraud, use **1930** and **cybercrime.gov.in** as the immediate reporting routes identified by I4C. Preserve messages, account details, payment references, URLs, and timestamps. Reporting and legal outcomes can depend on the facts and current official procedures, so follow the instructions provided by the relevant authority or service provider.
Limitations
- Evidence and guidance can change; verify current advisories and organisational applicability before acting.
Evidence sources
- cisa.gov — NSA, FBI, and CISA Release Cybersecurity Information Sheet on Deepfake Threats
- ic3.gov — FBI IC3: Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud
- ic3.gov — FBI IC3: Senior US Officials Impersonated in Malicious Messaging Campaign
- microsoft.com — Microsoft Cyber Signals Issue 9: AI-powered deception
- pib.gov.in — PIB/MeitY: India well-equipped to tackle evolving online harms and cyber crimes
- trends.google.com — Google Trends comparison: deepfake scam, AI voice scam, deepfake video scam