ChatGPT’s new image generator is insanely good at faking receipts — but OpenAI isn’t losing any sleep over it

Over the last several days, there’s been quite a buzz surrounding OpenAI, particularly following their reveal of a novel image generator tool for ChatGPT, which is driven by their GPT-40 model. Aren’t those Studio Ghibli memes neat?

ChatGPT’s image creation abilities appear remarkably advanced, surpassing simple meme production. Interestingly, it was found by our colleagues at TechRadar that this chatbot is extraordinarily proficient at creating convincing fake receipts.

Although the outlet acknowledges that AI-generated receipts may exhibit noticeable signs to distinguish them from genuine ones, it describes these receipts as being overly polished, with an almost digital or computer-generated appearance, and occasionally displaying errors in calculations.

No matter how carefully you examine a receipt, it’s possible to overlook certain signs. It’s also worth noting that digital editing tools such as Photoshop exist and can sometimes be used to rectify some apparent mistakes.

1. Even if you scrutinize a receipt, it may still contain subtle errors that go unnoticed. It’s important to remember that software like Photoshop is available for correcting certain obvious discrepancies.
2. A thorough examination of a receipt might not catch every potential error. Keep in mind that tools such as Photoshop can help rectify some apparent mistakes.
3. If you take a close look at a receipt, there’s still a chance you may miss certain signs of an issue. Don’t forget that digital editing software like Photoshop is available for correcting errors that might be visible to the naked eye.

According to TechRadar, the best method for creating a false receipt involves using ChatGPT as a tool to replicate an existing receipt. It’s interesting to note that you can modify the information on the receipt according to your specifications and requirements.

OpenAI says it’s not a big deal

Although this revelation could potentially pose a significant risk, OpenAI appears relatively unconcerned about digital con artists exploiting this situation for malicious purposes, as suggested by a representative from OpenAI who communicated with TechRadar.

We keep a close eye on generated images, both on and beyond our own system, ensure they are made using our software, and respond accordingly when we detect any breaches of our guidelines regarding their use.

Our continuous goal is to learn from practical applications and user responses, and we will persistently improve our rules to strike a balance between encouraging creativity and avoiding misuse. Every image carries standard C2PA metadata marking them as AI-produced by OpenAI.

As an analyst, I found it intriguing when OpenAI shared with TechCrunch that AI-generated receipts might not always be detrimental. Amidst this scenario, they suggested a creative approach: leveraging these fake receipts to enhance financial literacy education.

Read More

2025-04-03 12:09