AI fraud: Fake expense claims flood businesses

Artificial intelligence is playing an increasingly prominent role in tax and expense fraud. According to the Financial Times (paywall), verification platform AppZen now detects AI-generated forgeries in roughly 14% of all receipts – up from none just a year ago. Since the release of GPT-4o in May 2024, the number of manipulated invoices has surged.

Fraudsters are increasingly turning to AI image generators and text models to produce convincing fake invoices. With just a few prompts, logos, fonts and layouts can be replicated with striking accuracy. Even fine details such as watermarks can be generated automatically or lifted from real templates. Some tools even allow users to upload genuine receipts for the AI to modify – changing amounts, dates or other key information.

The fakes are so convincing that even seasoned auditors are often fooled. Where Photoshop skills were once required, today an AI tool and a few seconds are all it takes. Companies in Germany are feeling the impact especially hard: studies show that small and medium-sized businesses lose an average of €14,000 ($15,000) per year to expense fraud. Many perpetrators don’t seem to view it as a serious offense – according to an SAP survey, more than half of employees consider expense fraud of up to €100 ($110) acceptable.

Companies are now fighting AI with AI, using automated systems to analyze metadata and cross-check travel data. But these measures have their limits – a simple screenshot can erase all digital traces. On platforms like Reddit, many users take a pragmatic stance: “AI just makes fraud faster and cheaper – it’s not a new problem, just a new tool,” one comment read. Still, one thing is clear: anyone caught falsifying expenses risks losing their job – whether they used Photoshop or ChatGPT.