Security Stop-Press: AI Deepfake Receipts Spark Expense Fraud Concerns

Experts are warning AI-generated receipts are now so convincing that they could be used to cheat company expense systems.

OpenAI’s latest image generator, part of its ChatGPT 4o model, allows users to create fake receipts in seconds, complete with logos, stains, and creases. Online examples show how easily these images can bypass expense software, raising concerns among security experts.

For example, AI researcher Raphael Chenol has reported demonstrating how altering dates and prices on realistic-looking receipts now takes just seconds, where once it required graphic design skills. He warned that without safeguards, companies could soon face a flood of fraudulent claims.

Other security commentators say the risk isn’t limited to employee expense claims. It seems there are growing concerns that criminals could impersonate staff, submit fake receipts, and trick finance teams into making payments, particularly when paired with email-based scams targeting company accounts.

Although OpenAI says its images contain metadata showing they’re AI-generated, this can be removed. The company has defended the tool’s flexibility, saying it can also be used for education and creative work.

To reduce the risk, experts recommend multi-step approval processes and secure digital verification methods. As deepfake receipts become harder to spot, companies will need to rely less on visual checks, and more on trusted, auditable systems.