FBI / FTC / Interpol · 2025 · 03 · 15 · Incident · ~2 min read
AI voice clone scams hit billions in losses
What's actually new
- $12.5 billion in 2024 fraud losses in the US alone — driven heavily by AI voice cloning and deepfake calls.
- AI voice clones from seconds of audio. Free tools online imitate someone's voice from a 10-second clip, with results convincing enough to fool family members and bank tellers.
- Pig-butchering romance scams + deepfake video calls made traditional 'video chat to verify' useless. Old advice for spotting scams stopped working.
- Corporate impersonation fraud. Multiple multi-million-dollar wire transfers in 2024-2025 were authorised by finance teams who'd just been on a video call with what they thought was their CEO.
If you want more
Worth knowing
- Underreporting is severe. Embarrassment keeps victims silent — actual losses are likely much higher than reported.
- 'AI detection tools' are not a fix. The arms race favours generation, not detection. Don't rely on a tool to tell you a call is fake.
- Most laws weren't ready. By 2025-2026 a few US states had voice-cloning fraud statutes; most countries still relied on general fraud laws that hadn't been updated for AI.
Who should care
Anyone with elderly parents or grandparents — they're the highest-loss demographic. Anyone in a finance or executive-assistant role who handles wire transfers. Anyone on dating apps. Companies that pay invoices based on phone or email instructions. Parents whose teenagers' voices are on TikTok.
What to do about it
Set up a family safe word (see the Plain AI topic of the same name) — a single odd word that you and your loved ones use to confirm a real distress call. If a 'family member' calls in trouble and can't say the word, hang up. For business: require call-back to a known number for any wire-transfer instruction received via voice or video. Don't trust the caller ID. Don't trust the voice. Don't trust the video.
Honest take
Voice clone fraud is the AI story most commonly affecting people who have nothing to do with AI. The technology is simple, free, and unregulated; the people most vulnerable are the ones least equipped to suspect it. The slow-motion regulatory response is the failure here — by the time most countries pass meaningful laws, the playbook will be three generations old. The honest advice in 2025-2026: assume any unexpected voice call asking for money is a scam until proven otherwise. That's a depressing posture, but it's the safe one.
Sources
Last verified · 2026 · 05 · 05 · Found a fact wrong? corrections@aguidetocloud.com