Deepfake scams reach bank customers before they reach the bank

Published January 24, 2025 Updated October 28, 2026 

Deepfake social engineering scams use cloned voices, AI-generated video, and synthetic images to pressure people into sending money. In these scams the victim authorizes the payment, so fraud controls built to stop account takeover don’t fire. Financial institutions protect customers by warning them early, setting up a callback rule they’ll recognize, and scanning the suspicious media customers report.

These scams end with the customer pressing send. A grandmother answers a call in her grandson’s voice, asking for bail money tonight. The voice is cloned. The panic is scripted, and the transfer she makes looks normal to every fraud check her bank already runs.

How deepfake scams reach customers

Deloitte estimates generative AI-enabled fraud losses in the U.S. could reach $40B by 2027, up from $12.3B in 2023. Three patterns explain how these scams reach customers.

Family impersonation works on instinct. A caller sounds like a grandchild or a close friend, says they’re in trouble, and asks for money before anyone can check. Microsoft Research demonstrated voice cloning from a three-second audio sample. 

Investment scams borrow a famous face. AI-generated videos of well-known investors and business leaders promote returns that don’t exist, and they run as ads and posts in the same feeds customers scroll every day.

Romance scams take longer. A scammer uses AI-generated photos and video to build a relationship, then invents a crisis that only money can fix. 

Why these scams slip past fraud checks

Account-level fraud checks ask whether the person sending money is the real account holder. In a deepfake scam, they are.  The customer logs in, clears every check, and approves the payment because someone they trust seems to have asked. FinCEN warned financial institutions in November 2024 about fraud using deepfake media and AI-altered identity documents.

What financial institutions can do for customers

Institutions can’t screen every call a customer takes. They can make each call easier to see through. 

• Name the scripts in your warnings, so customers know an urgent call from a relative is never a reason to move money.
• Teach a callback habit, so customers hang up and call back on a number they already have before sending anything.
• Give customers a way to report what they received, so a fraud analyst can check the voice note, video, or image first.
• Pass new scam scripts from your fraud team to the people who answer customer calls and write customer warnings.

For detection inside contact centers and account opening, read how deepfakes target bank operations.

Where Reality Defender fits

Check the media before the money moves. When a customer forwards a suspicious voice note, video, or image, a fraud analyst can upload it to RealScan and get a result of Authentic, Suspicious, or Manipulated. Reality Defender detects AI-generated voice, video, and image for financial services teams. It doesn’t verify identity, so the result sits alongside the checks your team already runs.

Scams work when a customer acts before checking. Make checking the easy option.