OpenAI Model Bypassed to Generate Fraudulent Receipts and Prescriptions

# OpenAI Model Bypassed to Generate Fraudulent Receipts and Prescriptions

https://gcdradar.vercel.app/incidents/openai-model-bypassed-to-generate-fraudulent-receipts-and-pr · AIR-2025-0146 · closed 2025-04-14

- Lane: Research
- Severity: 1 (Research / proof of concept)
- Fix status: Fix status unknown
- First reported: 2025-03-31
- AI role: AI-enabled attack
- Products: ChatGPT

## Summary

Users bypassed safety restrictions in OpenAI's image generation model to produce fraudulent documents such as fake restaurant receipts and medical prescriptions. The generated material was used to demonstrate expense and medical forgery capabilities.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2025-03-31: OpenAI's 4o Model Allegedly Used to Generate Fake Receipts and Prescriptions (https://incidentdatabase.ai/cite/1018)

## Sources

- 2025-03-31 AI Incident Database: OpenAI's 4o Model Allegedly Used to Generate Fake Receipts and Prescriptions https://incidentdatabase.ai/cite/1018

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.