This month, ChatGPT unveiled a new image generator<\/a> as part of its 4o model that is a lot better at generating text inside images.\u00a0<\/p>\n People are already using it to generate fake restaurant receipts, potentially adding another tool to the already-extensive toolkit<\/a> of AI deepfakes used by fraudsters.<\/p>\n Prolific social media poster and VC Deedy Das posted on X<\/a> a photo of a fake receipt for a (real) San Francisco steakhouse that he says was created with 4o.<\/p>\n You can use 4o to generate fake receipts.<\/p>\n There are too many real world verification flows that rely on \u201creal images\u201d as proof. That era is over. pic.twitter.com\/9FORS1PWsb<\/a><\/p>\n \u2014 Deedy (@deedydas) March 29, 2025<\/a><\/p><\/blockquote>\n<\/div>\n<\/figure>\n Others were able to replicate similar results, including<\/a> one with food or drink stains to make it look even more authentic:<\/p>\n I think in the original image the letters are too perfect and they don\u2019t 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<\/a><\/p>\n \u2014 Michael Gofman (@michaelgofman) March 29, 2025<\/a><\/p><\/blockquote>\n<\/div>\n<\/figure>\n The most real-looking example TechCrunch found was actually from France, where a LinkedIn user posted<\/a> a crinkled-up AI-generated receipt for a local restaurant chain:<\/p>\n TechCrunch tested 4o and was also able to generate a fake receipt for an Applebee\u2019s in San Francisco:<\/p>\n But our attempt had a couple of dead giveaways that it was faked. For one, the total uses a comma instead of a period. For another, the math doesn\u2019t add up. LLMs still struggle to do basic math<\/a>, so this isn\u2019t particularly surprising.<\/p>\n But it wouldn\u2019t be hard for a fraudster to quickly fix a few of the numbers with either photo editing software or, possibly, more precise prompts.<\/p>\n It\u2019s clear that making it really easy to generate fake receipts presents huge opportunities for fraud. It wouldn\u2019t be hard to imagine this kind of tech being used by bad actors to get \u201creimbursed\u201d for entirely fake expenses.<\/p>\n OpenAI spokesperson Taya Christianson told TechCrunch that all of its images include metadata indicating they were made by ChatGPT. Christianson added that OpenAI \u201ctakes action\u201d when users violate its usage policies and that it\u2019s \u201calways learning\u201d from real-world use and feedback.<\/p>\n TechCrunch then asked why ChatGPT allows people to generate fake receipts in the first place, and whether this is in line with OpenAI\u2019s usage policies<\/a> (which ban fraud.)<\/p>\n Christianson replied that OpenAI\u2019s \u201cgoal is to give users as much creative freedom as possible\u201d and that fake AI receipts could be used in non-fraud situations like \u201cteaching people about financial literacy\u201d along with creating original art and product ads.<\/p>\n<\/div>\n\n
\n
![\"\"](\"https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/03\/image_e95a9d.png?w=453\")