AI scams: how to spot them and protect yourself

What AI changes

The playbook has not changed: urgency, fear, greed and trust. Generative AI simply makes the bait more convincing and cheaper to produce:

• The cloned voice: using text-to-speech technology, a scammer calls sounding like your child or partner: “I’ve been in an accident, send money now, don’t tell anyone.” The FTC warned about this exact scheme in March 2023 (FTC consumer alert).
• The fake video: a deepfake shows an executive on a video call asking for an urgent wire transfer, or a celebrity promoting a miracle investment.
• The perfect message: phishing emails and texts with no typos, personalized with your name and employer.
• The fake assistant: an app or website imitating a well-known AI assistant to harvest your login or card details.

Practical habits that work

1. Call back on a different channel. Hang up and dial the number you already have, never the one the caller gives you.
2. Agree on a family code word. A word or phrase only your close family knows, to ask for whenever a call feels off.
3. Treat unusual payment requests as red flags: gift cards, cryptocurrency, or an “urgent” wire to new bank details.
4. Never share a verification code sent by text, even with someone claiming to be your bank.
5. At work, require a second approval: no payment on the strength of a call or video alone, even from the boss.
6. Be mindful of what you post: videos with your voice, details about your family and travel plans.

If it happens to you

Don’t stay silent. Contact your bank right away, keep the evidence (numbers, messages, screenshots) and report it at ReportFraud.ftc.gov. Talking about it with friends and family also helps others spot the next attempt.

Common myths

• “I’d recognize my own family’s voice.” That trust is exactly what the scam exploits.
• “Bad grammar always gives scammers away.” Not since anyone can use a chatbot.
• “Only older people fall for this.” Businesses are targeted too, especially with fake-executive payment fraud.
• “An AI agent can’t be tricked.” Assistants that read your email or browse for you can themselves be manipulated by hidden instructions, a technique called prompt injection.