OpenAI’s newest replace to ChatGPT, launched earlier this month, has considerably improved the mannequin’s means to generate photographs containing reasonable textual content– a technical leap that’s already fuelling issues about new avenues for fraud.
Users have shortly found that the brand new GPT-4o mannequin can be utilized to generate convincing pretend receipts. What may as soon as have required Photoshop and a practiced hand can now be performed with a easy textual content immediate.
Among the primary to spotlight the chance was Deedy Das, a enterprise capitalist at Menlo Ventures. Posting on X, Das shared a picture of what seemed to be a receipt from a widely known eatery in San Francisco. In actuality, it had been generated fully by ChatGPT. “You can use 4o to generate fake receipts,” he wrote. “There are too many real world verification flows that rely on ‘real images’ as proof. That era is over.”
You can use 4o to generate pretend receipts.
There are too many actual world verification flows that depend on “real images” as proof. That period is over. pic.twitter.com/9FORS1PWsb
— Deedy (@deedydas) March 29, 2025
Other customers have gone additional, refining prompts or feeding actual payments into the mannequin to generate more and more authentic-looking receipts– the sort that may very well be used to submit pretend expense claims, create cast monetary information, or trick associates and employers into reimbursing cash that was by no means spent.
Tools for fraud at individuals’s fingertips
The ease with which AI can now produce fakes has deepened issues over how generative fashions could also be exploited. Fake receipts could seem trivial in comparison with deepfake movies or voice clones, however their potential for real-world monetary hurt is critical.
Fraudsters may, for instance, use such instruments to help pretend reimbursement requests in workplaces, create bogus documentation for tax deductions, or trick crowdfunding platforms with fabricated buy histories.
Online marketplaces, gig platforms, and journey reimbursement programs– lots of which depend on uploaded photographs of receipts as proof– may all be targets.
One can think about such functionality getting used to offer cowl for illicit transactions as properly.
What OpenAI stated
OpenAI says it’s conscious of the dangers and is actively monitoring the software’s use.
TechCrunch cited OpenAI spokesperson Taya Christianson as saying that each one photographs generated with ChatGPT embody metadata flagging them as AI-created. The firm, she added, “takes action” when customers violate utilization insurance policies and is “always learning” from real-world use and suggestions.
When requested why ChatGPT permits customers to generate pretend receipts within the first place– regardless of a transparent coverage ban on fraudulent exercise– Christianson stated OpenAI goals to offer “as much creative freedom as possible” to customers.
She prompt that receipt era may have professional functions, similar to academic instruments for monetary literacy, or inventive and promoting use.
