Businesses face a growing threat from employees using advanced AI tools to create realistic fake expense receipts. Expense management platforms report a sharp increase in such fraud attempts since the launch of improved image-generation models. Companies are now relying on AI detection systems to combat the deception.
The emergence of sophisticated image-generating AI models from companies like OpenAI and Google has made it easier for employees to submit fraudulent expense reports. According to expense software provider AppZen, AI-generated fake receipts accounted for about 14% of fraud attempts in September, up from none the previous year. Fintech firm Ramp reported flagging more than $1 million in fraudulent invoices within 90 days using its new software.
A survey by expense management platform Medius found that 30% of financial professionals in the US and UK observed a rise in falsified receipts following the release of OpenAI’s GPT-4o last year. The trend accelerated after OpenAI launched an improved image generation model for GPT-4o in March. "These receipts have become so good, we tell our customers, ‘do not trust your eyes,’" said Chris Juneau, senior vice-president and head of product marketing for SAP Concur, which conducts over 80 million AI-powered compliance checks monthly.
Previously, creating fake documents required photo-editing skills or paid services, but now free AI chatbots allow users to generate convincing receipts in seconds via simple text prompts. Examples shared by platforms show images with realistic wrinkles, detailed menu itemizations matching real venues, and even signatures. OpenAI stated it takes action against policy violations and includes metadata in its images indicating creation by ChatGPT.
"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," warned Sebastien Marchon, chief executive of Rydoo. To counter this, platforms like Ramp and SAP Concur use AI to scan image metadata for AI origins—though users can bypass this by photographing or screenshotting the images. Detection also involves checking contextual details, such as repeated server names or timestamps, and employee travel data.
"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," noted Calvin Lee, senior director of product management at Ramp. A July SAP study revealed that nearly 70% of chief financial officers suspect employees are using AI to falsify travel expenses or receipts, with 10% certain it has occurred. Mason Wilder, research director at the Association of Certified Fraud Examiners, described AI-generated fraudulent receipts as a “significant issue for organizations,” adding, “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.”