Edited IDs, receipts and claim photos get through the checks you run today. Resemble looks at whether the image was generated or altered, before you approve an account or release a payout.
A recent fraud audit at a large gig economy platform revealed that over 1 in 10 identity documents that passed their verification had been generated with AI.
Oboarding fraud detection
A real, stolen ID with a swapped photo clears document checks, liveness detection, and background screening. It then passes every future check, because the face and credential now match each other. That's how account opening fraud and synthetic identity fraud usually start.
Claims, refunds, and payouts
Receipts, damage photos, and proof of delivery get edited in seconds with a consumer image model and attached to a refund or a claim. This creates the same problem for insurance claim fraud detection and payment fraud detection: an edited image can be accepted as real, leading to a refund, claim approval, or payout.
Proof of income and address
Account takeover fraud often starts with one of these: payslips, bank statments and utility bills, edited to clear an income threshold or place someone at an address they have never lived at.
Repeat use fraud rings
Rings reuse one document template and one background, swapping only the face, so any single upload can look reasonable while the batch gives them away. Fake ID detection only helps here if it can see across submissions.
