Six months ago I posted on LinkedIn about how AI image generation capabilities would lead to an increase in fraud with expense claims in companies.

I wasn't surprised then over the weekend to see a headline in the FT ‘Do not trust your eyes’: AI generates surge in expense fraud’.

Finance teams are starting to find fake receipts mixed in with legitimate expense submissions, and the quality is good enough that most standard checks miss them completely. These aren't crude Photoshop jobs anymore. The receipts have perfect logos, realistic transaction numbers, and amounts that conveniently stay just under approval thresholds. They look legitimate because the AI tools generating them have learned from millions of real receipts, so they know exactly what a proper receipt should look like.

So what actually catches them? Some finance teams are getting smarter about cross-checking details. Does the type of merchant actually match what they're claiming? Does the time on the receipt make sense if the employee was supposedly in a client meeting at the time?

For larger claims, some companies are going back to the actual retailers or card processors to verify the transaction happened, rather than just trusting the image someone's uploaded.

Many large companies have sophisticated automated verification software for checking, but most SMEs rely on someone in their accounts department manually reviewing receipts, and they need to know this technology exists in the first place. The starting point is awareness - making sure your team knows that creating fake receipts with AI is now very easy, and that they need to look more carefully at what's being submitted.

Then there's the policy side. Employees need to understand that submitting fake receipts isn't just bending the rules, it's fraud. Gross misconduct at minimum, potentially illegal depending on how the law catches up with AI-generated documents. However, the ethical line is clear - it's dishonest, and it undermines trust across the entire organisation.

Practically speaking, finance teams can start doing more manual spot checks on the details. It's tedious work, but for businesses without automated tools, a bit of extra scrutiny on higher-value claims or patterns that seem off can catch problems before they become bigger issues. Sometimes the best defence is just making it known that someone's actually paying attention.