Analysis · curated 11 Sep 2026

Prompt Injection via OCR: How a Facsimile Document Fools AI Identity Verification

First reported today pasqualepillitteri.it

Coverage timeline

Single-source analysis — first reported, latest, and curated coincide.

Why it matters

Prompt injection via OCR against AI identity-verification systems shows attackers can bypass automated KYC checks by embedding instructions in document images, undermining a security control many organizations now trust.

Article analyzes how a facsimile (fake) document can pass an LLM-based identity verification (KYC) check via prompt injection through OCR, drawing on a CERT-AGID case involving INPS-themed smishing in which AI-based document verification was deployed. It connects the technique to established multimodal/visual prompt injection research (Simon Willison, OWASP LLM01).