Workers are using AI tools to create fake expense receipts, companies warn

Companies across the globe are facing a growing problem as workers use artificial intelligence software to create fake expense receipts, marking a high-tech twist on workplace fraud.

The problem has worsened since major technology companies, such as OpenAI and Google, released new AI programs capable of creating realistic images. Business expense software companies say they are now seeing many more fake receipts being turned in by employees.

AppZen, which makes software to track business expenses, found that fake receipts made by AI represented about 14 percent of all fraudulent papers submitted in September. Last year, there were none. Another company, Ramp, said its new detection software caught more than $1 million worth of fake invoices in just 90 days.

A survey of money managers in the United States and the United Kingdom showed the scope of the issue. About 30 percent of financial workers questioned by expense management company Medius said they had noticed more fake receipts after OpenAI released GPT-4o last year.

“These receipts have become so good, we tell our customers, ‘do not trust your eyes’,” said Chris Juneau, who works as senior vice-president and head of product marketing for SAP Concur. His company is one of the largest expense tracking platforms in the world, processing more than 80 million compliance checks each month using AI.

Many platforms noticed a big jump in AI-made receipts after OpenAI launched GPT-4o’s better image creation features in March.

When asked about the issue, OpenAI told the Financial Times that it takes steps when people break its rules. The company also said its images contain hidden information that shows they were made by ChatGPT.

AI makes document fraud accessible to everyone

In the past, making fake receipts required knowing how to edit photos or paying someone else to do it through online services. Now, free and easy-to-use AI programs let workers create false receipts in seconds just by typing simple instructions to computer chatbots.

Several fake receipts shown to the Financial Times by expense management companies looked very real. They included wrinkled paper, detailed lists of items that matched actual restaurant menus, and signatures.

“This isn’t a future threat; it’s already happening. While currently only a small percentage of non-compliant receipts are AI-generated, this is only going to grow,” said Sebastien Marchon, who runs expense management platform Rydoo.

Companies fight back with AI detection tools

Because these fake receipts appear so realistic, companies are now using AI programs to detect them, as human workers struggle to distinguish between the real and fake.

The detection software works by checking the hidden data in image files to see if an AI program created them. But workers can easily remove this information by taking a photo or screenshot of the fake receipt.

To fight back, the software also examines other details, such as repeated server names and times, as well as broader information about the worker’s business trip.

“The tech can look at everything with high details of focus and attention that humans, after a period of time, things fall through the cracks, they are human,” said Calvin Lee, who works as senior director of product management at Ramp.

Research conducted by SAP in July found that nearly 70 percent of chief financial officers believed their employees were using AI to attempt to falsify travel expenses or receipts. About 10 percent said they were certain it had happened at their company.

Mason Wilder, who works as research director at the Association of Certified Fraud Examiners, called AI-generated fake receipts a “significant issue for organisations.”

He said: “There is zero barrier for entry for people to do this. You don’t need any kind of technological skills or aptitude like you maybe would have needed five years ago using Photoshop.”

Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.