ChatGPT can now generate fake receipts realistic enough to scam businesses

[zohaibahd]

Facepalm: When ChatGPT isn't being used to generate those viral Studio Ghibli-inspired images flooding your feeds, it can be tapped to fake receipts instead. The AI chatbot is finally decent enough at comprehending and generating text within images that crafty users realized it could produce fake receipts. These could then easily pass as legit proof of purchase to an unsuspecting set of eyes.

Until recently, AI-generated images had one glaring weakness – garbled or nonsensical text. Many models have managed to fix problems like messed-up hands to the point where they're not even an issue anymore. But text has remained a bit of an Achilles' heel. That's changing fast, though.

For instance, following the latest update to ChatGPT's image generation on its 4o model, users can spit out receipts with near-perfect formatting, itemized lists, and even realistic-looking logos. Of course, they're still not completely flawless and can look overly crisp, have text with odd spacing, and even have math errors.

I think in the original image the letters are too perfect and they don't bend with the paper. They look like hovering above the paper. Here is my attempt to make it more realistic. Let me know what you think. pic.twitter.com/EixRSHubeY

– Michael Gofman (@michaelgofman) March 29, 2025

However, receipts – especially for smaller amounts – aren't often scrutinized too heavily. A few extra tweaks in Photoshop, like adding stains, could easily fool an employer. The payoff is obvious to scammers. All they'd need to do is spend just a few minutes using ChatGPT and Photoshop, and even if the trick only works occasionally, it's still a low-effort, high-reward scheme.

Right now, the most effective method might be to feed an existing receipt to ChatGPT and modify details like prices or dates. But as the tech improves, entirely fabricated receipts could become indistinguishable from real ones.

You'd think that all this would ring alarm bells at OpenAI HQ, but while the company has acknowledged the risk, it isn't hitting the panic button just yet. A company spokesperson explained to TechRadar that it tracks image generation both on and off its platform, uses internal tools to verify AI-created content, and takes action when usage policy violations occur.

Interestingly, OpenAI also sees an upside to this capability. Spokesperson Taya Christianson suggested to TechCrunch that fake receipts could be used for financial literacy education by teaching people how to track expenses or spot fraud in a controlled setting.

Still, one has to wonder if this capability results in a net positive for society.

Permalink to story:

ChatGPT can now generate fake receipts realistic enough to scam businesses

 

[ScottSoapbox]

Most expense accounts have enough guard rails to make this a nonissue. Maybe you skirt a no alcohol rule but you are talking hundreds of dollars of petty cash at risk here. You can't even buy a GPU with that these days.

People with bigger expense accounts make enough that stealing a dinner isn't even worth their time to do it.

 

[The Talking Tech]

And sadly under the Trump administration expect no legislation to hold accountable the companies behind the A.I technology.

 

[The Talking Tech]

Most expense accounts have enough guard rails to make this a nonissue. Maybe you skirt a no alcohol rule but you are talking hundreds of dollars of petty cash at risk here. You can't even buy a GPU with that these days.

People with bigger expense accounts make enough that stealing a dinner isn't even worth their time to do it.

The fact that it's possible doesn't mean it's not "worth it"... stealing even a pen from an office store is still considered a crime no matter how insignificant its value.

The real issue here is that there no guard rails under the current administration or serious plans for legislation to keep companies in check, all it takes is the United States to started it for other countries to follow up with their own.

 

[wiyosaya]

Still, one has to wonder if this capability results in a net positive for society.

Seriously?

Maybe I'm wrong, but anyone using this tech, in any capacity, has the financial resources to actually pay for their meals or whatever else they are putting into their stomach(s).

Personally, I cannot think of any ethical use for this in any context. Are you suggesting that AI helping people grift results in a net positive for society??

IMO, there are enough people grifting in the world today. In no way does AI helping people grift result in a "net positive" for society.

 

[ScottSoapbox]

The fact that it's possible doesn't mean it's not "worth it"... stealing even a pen from an office store is still considered a crime no matter how insignificant its value.

The real issue here is that there no guard rails under the current administration or serious plans for legislation to keep companies in check, all it takes is the United States to started it for other countries to follow up with their own.

I've always disliked the stealing a pen argument. Being concerned over office supplies in an age where we work at home is just silly not "it's the principal" leadership. Should I bill my company for the electricity I used working late into the night or the printer paper I used because the supplier wanted a wet ink signature? How many tens of thousands of dollars in recruiting and training should we waste because Bill took a pen for his work laptop bag?

It is not the governments responsibility to put guardrails on companies reimbursement policies.

It is not the US governments responsibility to lead other countries governments by the nose.

Most companies have corporate cards so at most you are changing what you bought at an approved vendor (Thus my maybe you rename your alcohol comment). AI companies should be held to task for things like stealing everyone's content, but this receipt things is only upsetting to people that don't have expense accounts.

 

[wiyosaya]

Most companies have corporate cards so at most you are changing what you bought at an approved vendor (Thus my maybe you rename your alcohol comment). AI companies should be held to task for things like stealing everyone's content, but this receipt things is only upsetting to people that don't have expense accounts.

So if a restaurant is a victim of a fake receipt scam, or if a company is handed a fake receipt for reimbursement for something that their employee never paid for, are you saying these are examples of victimless crimes?

I think your argument amounts to nothing more than a red-herring.

Given the number of people that don't GAS about how their actions affect the finances of others, I think that without laws against stealing, even petty amounts, instances of stealing would be far worse, and if people are incapable of policing themselves,where, then, does social responsibility for policing lie?

When is the monetary value of the loss meaningful?

Many instances of credit card fraud these days are for small dollar values because so many people do not track their credit card expenses, and, therefore,such fraud goes unnoticed. Yet those instances are still fraud and add up over time - with the likely result that the card-holder reports those instances to their bank, and their bank reimburses them - thus making the bank the victim of the fraud.

 

[GodisanAtheist]

Literally anything that relies on single factor authentication is basically going to be obsolete thanks to AI (and probably most things that have basic two factor authentication as well, thinking like a colored wristband with a bar/QR code).

Its really remarkable how *not ready for this* our world is and how much tech is still 1970's "Just make it hard to reproduce and that's enough security" so many things are.

 

[ScottSoapbox]

So if a restaurant is a victim of a fake receipt scam, or if a company is handed a fake receipt for reimbursement for something that their employee never paid for, are you saying these are examples of victimless crimes?

I think your argument amounts to nothing more than a red-herring.

Given the number of people that don't GAS about how their actions affect the finances of others, I think that without laws against stealing, even petty amounts, instances of stealing would be far worse, and if people are incapable of policing themselves,where, then, does social responsibility for policing lie?

When is the monetary value of the loss meaningful?

Many instances of credit card fraud these days are for small dollar values because so many people do not track their credit card expenses, and, therefore,such fraud goes unnoticed. Yet those instances are still fraud and add up over time - with the likely result that the card-holder reports those instances to their bank, and their bank reimburses them - thus making the bank the victim of the fraud.

I don't believe you understand a red-herring. Or read my arguments. You certainly aren't addressing them or staying on topic. -dramatic pose- What is justice? Morality? Meaningful dialog? -end scene-

 

[OortCloud]

Surely tills are able to tie receipts back via a transaction id to check a receipt is valid?

 

[Vanderlinde]

You always need a matching reference, one that is registered at the local restaurant or whatever with it's own reciept ID.

ChatGTP can't fake that.

 

[kiwigraeme]

You always need a matching reference, one that is registered at the local restaurant or whatever with it's own reciept ID.

ChatGTP can't fake that.

Why not, some companies spit out thousands of receipts , seen lots of different
receipts in my time . pretty sure for low value stuff customer services just do a vey casual check .


In my county a lot of bg appliance stores kept a digital receipt under your phone number . stops hassle of lost or fade to nothing in 4 months

 

[kiwigraeme]

What to see some gaslighting
" OpenAI also sees an upside to this capability. Spokesperson Taya Christianson suggested to TechCrunch that fake receipts could be used for financial literacy education by teaching people how to track expenses or spot fraud in a controlled setting"

That's some fine corporate dribble right there

Heres some STLs to print lethal 3D weapons , so you know how to spot them in the wild !!

 

[Theinsanegamer]

And sadly under the Trump administration expect no legislation to hold accountable the companies behind the A.I technology.

The fact that it's possible doesn't mean it's not "worth it"... stealing even a pen from an office store is still considered a crime no matter how insignificant its value.

The real issue here is that there no guard rails under the current administration or serious plans for legislation to keep companies in check, all it takes is the United States to started it for other countries to follow up with their own.

The problem with this line of argumentation, of course, is that the last administration had no interest in guardrails either. I seem to remember, however, that OpenAI was pursuing legislation (that thankfully got axed) that would have given them a de facto monopoly on AI activities in the USA. Kinda like how Sam Bankman was pursuing legislation that would have given his company the power to regulate crypto.....

Also, it is not the US' job to lead the world. Worldwide many countries have made it clear they do not like the US nor do they want the US to make decisions for others anymore.

I've always disliked the stealing a pen argument. Being concerned over office supplies in an age where we work at home is just silly not "it's the principal" leadership. Should I bill my company for the electricity I used working late into the night or the printer paper I used because the supplier wanted a wet ink signature? How many tens of thousands of dollars in recruiting and training should we waste because Bill took a pen for his work laptop bag?

It is not the governments responsibility to put guardrails on companies reimbursement policies.

It is not the US governments responsibility to lead other countries governments by the nose.

Most companies have corporate cards so at most you are changing what you bought at an approved vendor (Thus my maybe you rename your alcohol comment). AI companies should be held to task for things like stealing everyone's content, but this receipt things is only upsetting to people that don't have expense accounts.

Stealing is still stealing. Just because you work from home doesnt mean theft doesnt count anymore. Yes, it's leadership that is petty, but this is "two wrongs dont make a right" territory. If you work from home, and the company demands you be in front of your computer with monitoring software, then IMO you should be able to charge for energy and write the space off on your taxes.

It's not the governments responsibility to regulate return policies (although some argue it should be) but making fake receipts and using them to get money is fraud, and this is absolutely under the purview of the government.