Dossier: AI Detection of Forged Applicant Documents
Krystian Siek
Dossier checks applicant documents for forgery, for lenders, letting agents and AML teams.
What it does
Examines bank statements, utility bills, payslips and IDs inside the file: edits after issue, hidden text and retyped amounts.
Checks balances row by row and passport MRZ check digits.
Compares names, addresses and dates across the documents and against the applicant's declaration.
Every finding is explained and decided in code, not by the model.
Published eval (synthetic test packs)
40 of 40 forged packs caught, 0 false alarms on the genuine packs.
Stack: Python, TypeScript, PostgreSQL, Docker.
Try it, forge an applicant yourself: https://adaptivesystems.dev/dossier/try
Like this project
Posted Nov 5, 2026
Checks bank statements, bills, payslips and IDs for edits after issue, hidden text and retyped amounts, for lenders, letting agents and AML. Published eval: 40/40 forged packs caught, 0 false alarms.
Likes
0
Views
3
