Topics
Latest
AI
Amazon
Image Credits:Kim Jae-Hwan/SOPA Images/LightRocket / Getty Images
Apps
Biotech & Health
Climate
Image Credits:Kim Jae-Hwan/SOPA Images/LightRocket / Getty Images
Cloud Computing
DoC
Crypto
Image Credits:chagpt / charles rollet
Enterprise
EVs
Fintech
Image Credits:chagpt / charles rollet
fund raise
contraption
Gaming
Government & Policy
Hardware
layoff
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
Security
Social
Space
Startups
TikTok
Transportation
Venture
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
This calendar month , ChatGPT unveil a raw image generatoras part of its 4o theoretical account that is a lot better at generating textual matter inside epitome .
hoi polloi are already using it to get bastard eating house receipts , potentially adding another tool to thealready - extended toolkitof AI deepfakes used by fraudsters .
Prolific societal media poster and VC Deedy Dasposted on Xa pic of a fake receipt for a ( actual ) San Francisco steakhouse that he says was created with 4o .
you could use 4o to generate phoney gross . There are too many real world verification flows that rely on “ real images ” as proof . That era is over.pic.twitter.com/9FORS1PWsb
Others were capable to replicate alike answer , includingone with solid food or salute grunge to make it look even more authentic :
I think in the original image the letters are too thoroughgoing and they do n’t bend with the report . They attend like vacillate above the paper . Here is my attempt to make it more naturalistic . get me know what you think.pic.twitter.com/EixRSHubeY
The most real - looking example TechCrunch establish was really from France , where a LinkedIn userposteda crinkled - up AI - bring forth receipt for a local eating house chain :
TechCrunch tested 4o and was also able to generate a fake receipt for an Applebee ’s in San Francisco :
But our endeavor had a brace of dead giveaways that it was faked . For one , the sum uses a comma rather of a point . For another , the math does n’t tot up up . LLMsstill skin to do canonic math , so this is n’t peculiarly surprising .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
But it would n’t be hard for a fraudster to quickly fix a few of the act with either photo editing software or , possibly , more accurate prompting .
It ’s cleared that stimulate it really easy to engender fake revenue presents huge chance for pseudo . It would n’t be hard to ideate this kind of tech being used by bad actors to get “ reimbursed ” for entirely fake expense .
OpenAI spokesperson Taya Christianson separate TechCrunch that all of its range include metadata indicating they were made by ChatGPT . Christianson add together that OpenAI “ take action at law ” when users violate its exercise policies and that it ’s “ always learning ” from real - world exercise and feedback .
TechCrunch then asked why ChatGPT permit mass to generate simulated receipts in the first topographic point , and whether this is in line with OpenAI’susage policies(which prohibition fraud . )
Christianson replied that OpenAI ’s “ end is to give drug user as much creative exemption as potential ” and that phony AI receipts could be used in non - fraud situations like “ teaching people about financial literacy ” along with make original artistry and ware advert .
