AI-enabled fraud is 4.5 times as lucrative, and the scam-compound model is getting smaller

The Financial Action Task Force’s president, Giles Thomson, warns that AI is changing the economics of industrial fraud. Small teams can now create persuasive fake websites, impersonate targets across languages and use deepfake audio or video in cryptocurrency, investment and romance scams — capabilities that previously required larger organizations. Interpol estimates that AI-enabled fraud is 4.5 times as lucrative as traditional scams; the FBI recorded nearly $893mn in related US losses last year, while estimated global fraud losses reached $442bn.

That does not mean the Southeast Asian scam-compound system has disappeared. Forced labor, trafficking and physical compounds remain central to many operations. AI makes the model more productive and portable: fewer operators can target more victims with more personalized approaches, complicating enforcement focused on conspicuous physical sites.

FATF plans to recommend national anti-scam centers that connect banks, technology platforms and law-enforcement agencies. Coordination matters because each participant sees only a fragment — a suspicious payment, a spoofed account or a victim report. AI can also aid defense by detecting synthetic media and allowing investigators to engage suspected scammers at scale, creating an arms race rather than a one-way advantage.

Fraud policy designed around large call centers and after-the-fact money tracing is falling behind an industry that can now generate credible identities, relationships and sales pitches almost instantly.