Deepfake scam
A deepfake scam is a fraud scheme that uses AI-generated audio, images, or video to impersonate a real person — an executive, a family member, a celebrity, or an official — so the victim trusts a request they would otherwise question. Voice-clone calls, fake video meetings, and synthetic celebrity endorsements are the dominant forms.
A deepfake scam is ordinary fraud with an extraordinary upgrade: the scammer no longer has to claim to be someone you trust — they can sound and look like them. AI voice cloning and face synthesis turned the weakest part of every con (believability) into its strongest, and the loss data shows it: the FBI's IC3 logged $12.5 billion in reported internet-crime losses in 2024, Sumsub measured a 245% year-over-year rise in deepfake-enabled fraud attempts, and Mastercard found 46% of businesses have already been targeted.
The five dominant deepfake scam types
1. Executive impersonation (CEO fraud)
A cloned executive voice — or a fully synthetic video meeting — orders an urgent, confidential transfer. The defining case is Arup in Hong Kong: 15 transfers, $25.6 million, every meeting participant except the victim a deepfake. Ferrari narrowly escaped the same play when the target quizzed the fake CEO on something only the real one knew.
2. Family-emergency voice clones
"Grandma, I'm in trouble — don't tell Mom." A 30-second public clip is enough to clone a grandchild's voice. This is the highest-volume consumer pattern; the vishing examples page breaks down the full script.
3. Celebrity endorsement scams
Synthetic video of a celebrity pitching a crypto platform, miracle product, or giveaway. Music companies alone are fighting this at industrial scale — Sony reported 135,000+ takedown requests for AI fakes of its artists.
4. Romance and identity scams
Long-running personas backed by AI-generated profile photos and, increasingly, live face-swapped video calls that "prove" the persona is real.
5. Political disinformation with a fraud tail
Cloned public figures push fake instructions or investment schemes — the New Hampshire Biden robocall told voters to stay home and drew a $6M FCC fine, which also established that AI-voice robocalls violate the TCPA.
Anatomy of every deepfake scam
Strip away the medium and each scam has the same skeleton — the same one described in the vishing attack breakdown:
| Stage | What happens | Your countermove |
|---|---|---|
| Reconnaissance | Scammer collects voice/face samples from public content | Limit long public clips of decision-makers where practical |
| Impersonation | Voice clone or face swap + spoofed caller ID/account | Assume identity ≠ channel: the voice proves nothing |
| Urgency | A deadline that forbids verification | Urgency itself is the alarm signal |
| Extraction | Wire, gift cards, crypto, or one-time codes | Irreversible payment method = stop |
How to check if you're being scammed
- Verify out-of-band. Hang up. Call the real person or institution on a number you already have. This defeats every clone and every spoofed caller ID, every time.
- Test the media. Voicemail or voice note → free AI voice detector. Photo → AI image detector. Video clip → deepfake video detector. Each returns a verdict with the reasoning behind it, no signup.
- Screen the script. The scam call screener matches a recording against 20 known scam-script categories.
- Study real cases. The Deepfake Incident Database documents hundreds of verified attacks with sources — the fastest way to calibrate your instincts is to read ten of them.
Organizations facing this at scale — banks, insurers, call centers — run the same detection continuously via API; see the banking and call-center playbooks.
Frequently asked questions
What is a deepfake scam? Fraud that uses AI-generated voices, images, or video to impersonate someone the victim trusts — a boss, a family member, a celebrity, or an official.
What's the biggest deepfake scam on record? Arup's $25.6M loss to a fully deepfaked video meeting.
How do I know if I'm being scammed? Urgency + an irreversible ask is the signature. Verify out-of-band and run any recording through the deepfake detector.
Can deepfake scams be detected? Yes — detection models catch voice-clone and face-synthesis artifacts in recordings, and enterprises run them in real time on calls and meetings.
Related reading
- Deepfake fraud — the enterprise view: losses, liability, controls
- Vishing · Vishing examples · Vishing attack
- How to spot a deepfake scam call
- What is a deepfake?