AI-Generated Fake Receipts Used in Workplace Expense Fraud
Employees reportedly are using AI image generation tools to create realistic fake expense receipts for fraudulent reimbursement claims. Expense management platforms reported a sharp increase in detected AI-generated fraudulent documents following the release of improved image generation models in 2024-2025.
What happened
Employees are submitting AI-generated fake expense receipts to their employers. Software provider AppZen reported that fake AI receipts accounted for roughly 14% of all fraudulent documents submitted last month, compared to zero last year.
Expense management platform Ramp caught more than $1 million in fraudulent invoices within 90 days. Around 30% of British and American financial professionals surveyed by expense management platform Medius reported seeing an uptick in falsified receipts after the launch of OpenAI's GPT-4o in 2024.
A significant jump in AI-generated receipts followed the debut of GPT-4o's improved image generation model in March. The receipts feature realistic details including wrinkles in paper, detailed itemization in line with real-life menus, and signatures.
Reported impact
Businesses are experiencing increased fraud from AI-generated fake receipts submitted through expense reports. The accessibility of free image generation tools allows employees to quickly falsify receipts in seconds using simple text commands to chatbots.
Index analysis
The rise of image generation models from companies like Google and OpenAI has led to a wave of AI-generated receipts submitted in-house. SAP Concur processes more than 80 million compliance checks per month using AI and warned that the receipts have become so realistic that 'Do not trust your eyes.'
Governance controls
OpenAI stated that it takes action when its policies are violated and its images contain metadata that indicate they were created using ChatGPT. Expense management platforms have implemented detection systems to identify fraudulent documents, though the realism of AI-generated receipts makes detection challenging.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.