1. AI voice-clone phone scams
A voice-clone scam begins with a small sample of someone’s speech: a public video, a voicemail greeting, or a clip shared on social media. Software can use that sample to create a voice that sounds close enough to a relative, colleague, or supervisor to create panic in a short phone call.
The script is familiar: a loved one is in trouble, a payment must be made immediately, and there is a reason not to tell anyone else. The AI component changes the emotional pressure. Hearing a recognizable voice can make an urgent story feel true before the listener has time to question it.
2. Deepfake video and identity verification
Some financial services, employers, and online platforms ask people to show their face on camera as part of identity verification. Criminals can use manipulated video, stolen identity data, and synthetic images in an attempt to pass those checks or to persuade a real person that they are speaking with a legitimate representative.
A polished video is not proof of identity. Lighting, audio, and facial expressions can now be altered or generated quickly. This does not mean every video call is fraudulent; it means an unexpected request to change account details, approve access, or send money deserves an independent check.
3. AI-personalized phishing emails
Older phishing messages often had obvious spelling mistakes or generic greetings. AI can help criminals draft clear, natural-sounding messages and tailor them with details gathered from public profiles, breached data, or a company website. A message may mention a real manager, a recent purchase, or a local event.
The goal is still the same: get you to click a link, disclose a password, share a one-time code, or move money. Personal details can make an email feel credible, but they do not prove the sender controls the account they claim to represent.
4. Why old advice fails now — and what to do instead
“Look for bad grammar” is no longer enough. AI can improve a scammer’s wording, translate it fluently, and generate convincing audio or video. The strongest defense is not deciding whether a message looks polished. It is changing how you verify a request.
Use a separate, known channel. Hang up and call a family member using a saved number. Sign in through the official app or a web address you type yourself, rather than a link in the message. Contact your employer, bank, or service provider through an established number before acting on an unexpected request. The FBI’s Internet Crime Complaint Center reported 22,364 AI-related complaints and $893 million in reported losses in its 2025 data. The technology is evolving, but this verification habit remains practical and effective.