Scam Losses Estimated at $442 Billion: How AI Changes the Fraud Playbook

Online scams have grown into a criminal business measured in hundreds of billions of dollars.

The Global Anti-Scam Alliance’s (GASA) 2025 report estimates $442 billion in scam losses across 42 surveyed countries during the 12 months preceding its March 2025 survey. Separately, GASA says scammers are adopting AI to make their operations more convincing, targeted, and scalable.

For banks, payment companies, and online platforms, the challenge is identifying deception before a convincing conversation becomes a costly payment.

AI can automate more of a scam operation

GASA's 2026 policy agenda calls for intervention before victims send money, with financial services and technology platforms expected to play a larger role in prevention.

A separate GASA analysis of AI-powered fraud says scammers are using AI to scale operations and make interactions more convincing and targeted. Generative tools can assist with language, impersonation, and content creation, reducing some of the manual work behind each approach.

Voice impersonation brings the risk into phone calls. AI voice-cloning scams can imitate someone a victim knows, making familiar audio unreliable as proof of identity. Face cloning over video calls can do the same on screen, making visual confirmation unreliable too. 

Automation still leaves other parts of the fraud operation intact. Scam groups need ways to collect money, move the proceeds, and keep their infrastructure running after a victim engages.

Scam networks were lucrative before the latest AI tools

Global Financial Integrity's Shadow Figures report cites an earlier estimate that scam operations in Cambodia, Laos, and Myanmar steal about $43.8 billion annually, equivalent to roughly 40% of the countries’ combined GDP. Regional activity includes investment fraud and romance schemes, with GFI also connecting the scam economy to human trafficking and organized crime.

Separately, a $431,000 AI deepfake romance scam illustrates how impersonation can lead to substantial losses for an individual victim.

Proceeds can move through banks and cryptocurrency platforms, with transfer apps or gift cards providing other routes across jurisdictions. This movement can complicate recovery when stolen money crosses several financial systems before reaching its destination. 

What eWeek found: Older Americans account for nearly 40% of AI-related losses

eWeek analyzed the FBI's 2025 Internet Crime Complaint Center report and found a pronounced age imbalance in complaints containing AI-related information.

One in seven AI-related complaints came from people 60 and older, yet the group accounted for nearly two in five dollars lost. Average loss per complaint topped $112,000, almost three times the overall figure. IC3 classifies these complaints as AI related because AI appears in the reported incident, although the category does not assign every dollar lost specifically to AI.

The comparison shows substantially higher average losses in complaints from people 60 and older. It does not establish whether AI increased those losses or whether a few unusually costly cases drove the difference. For fraud prevention teams, the findings support examining potential loss size alongside complaint volume when assessing payment risks.

Volume tells you how often fraud is appearing. Loss size tells you where the financial damage is concentrated.

If you run fraud controls at a bank, payment company, telecom provider, or online platform, tighten checks around high-value payments.

1. Recheck first-time high-dollar recipients. Compare the transfer with the customer’s normal payment history and confirm the request through a separate channel.
2. Do not rely on voice or video alone. AI voice-cloning scams can imitate someone a customer trusts, so familiar audio or a recognizable face should not be enough to approve a payment.
3. Escalate sudden changes in payment behavior. When a customer moves a six-figure sum after weeks of online contact or a new relationship, pause the transfer and verify who requested it and why.

Teams can also track prevented loss, as well as how many suspicious high-value transfers are stopped before settlement and how quickly funds are recovered after a bad payment. Fraud prevention teams can then judge whether their controls are actually preventing the most costly losses.

More AI news: An Anthropic test went off course when Claude treated a real police cold-case form like a practice task and submitted false information.