GPT-4o generates realistic fake receipts enabling fraud schemes
OpenAI's GPT-4o image generator reportedly produces hyper-realistic fake receipts that could be used to commit fraud through expense reimbursement scams and other financial deception. The capability was demonstrated in March 2025 by users who shared examples and techniques for generating nearly authentic-looking receipts, though OpenAI states that generated images include identifying metadata.
What happened
OpenAI's GPT-4o model, released in March 2025 with an improved image generator, can produce hyper-realistic fake receipts with text, wrinkles, stains, and background details that are nearly indistinguishable from genuine ones.
Examples include fabricated receipts for real restaurants like San Francisco's Epic Steakhouse and Applebee's, complete with itemised purchases and totals. Users shared prompt templates and refined outputs to enhance authenticity, though minor errors such as comma-decimal confusion and arithmetic mistakes persist.
Reported impact
The capability enables fraudulent expense reimbursements, bogus tax deductions, fabricated crowdfunding claims, and forged financial records. Systems relying on receipts for verification—such as workplace reimbursement platforms, gig economy apps, and online marketplaces—are considered particularly vulnerable to fraud.
Trust in digital verification systems may erode, and businesses could face increased fraud costs while individuals could be deceived into reimbursing fake expenses or donating to scams.
Index analysis
The improved text-rendering capability in GPT-4o was intended for creative applications like educational content or advertising. Users quickly exploited this feature for malicious purposes, and minor errors can be easily corrected using basic photo editing software or more precise prompts.
Governance controls
OpenAI stated that all generated images include metadata identifying their AI origin and that policy violations are addressed. However, critics argue this fails to mitigate risks, as metadata can be stripped or ignored in manual review processes.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.