By asahifile AI-generated Receipts Are Making Expense Fraud Easier for Employees
In recent months, the emergence of new image-generation models by companies like OpenAI and Google has enabled employees to easily create fake expense receipts, leading to a notable increase in fraud. According to expense software provider AppZen, approximately 14% of fraudulent expense documents submitted in September were AI-generated, a dramatic rise from zero last year. Additionally, fintech company Ramp reported that within just 90 days, its software flagged over $1 million in fraudulent invoices.
A survey conducted among financial professionals in the US and UK by expense management platform Medius revealed that nearly one-third reported an uptick in falsified receipts since the introduction of OpenAI’s GPT-4o last year. Chris Juneau, senior vice-president at SAP Concur, noted, “These receipts have become so good, we tell our customers, ‘do not trust your eyes.’”
The launch of GPT-4o’s upgraded image generation model in March has amplified the volume of such fraudulent submissions. OpenAI emphasized that they actively respond to policy violations, utilizing metadata to identify images created by their tools.
Previously, crafting fraudulent documents required advanced photo editing skills, but now, employees can produce convincing receipts in seconds by providing simple instructions to AI. Some receipts shared with the Financial Times exhibited impressive realism, featuring accurate itemization that mimics actual menus, authentic signatures, and even wrinkles in the paper.
Sebastien Marchon, CEO of expense platform Rydoo, warned, “This isn’t a future threat; it’s already happening,” underscoring the likelihood of growth in AI-generated fraudulent receipts. The increasing sophistication of these counterfeits has prompted companies to leverage AI themselves for detection, as human reviewers often cannot identify such high-quality fakes.
Expense management software now analyzes receipts for image metadata to detect AI involvement, though users can easily circumvent this by photographing or screenshotting the receipts. Advanced algorithms are also employed, examining factors such as repeated names and timestamps to detect inconsistencies.
Research by SAP indicates that nearly 70% of CFOs suspect employees are attempting to falsify travel expenses or receipts using AI, with about 10% certain that it has occurred within their organization. Mason Wilder, research director at the Association of Certified Fraud Examiners, called this trend a “significant issue.”
He pointed out that anyone can produce these frauds without the technological expertise previously necessary, further exacerbating the problem.
As companies grapple with this burgeoning issue, the demand for solutions to combat AI-generated fraud is likely to rise, making it crucial for expense platforms to stay ahead of the curve in an increasingly digital landscape.