ChatGPT’s Enhanced Image Generator Raises Concerns Over Fake Receipts

So, ChatGPT just rolled out this fancy image generator with its 4o model, and guess what? It’s gotten scarily good at putting text in images. Like, so good that people are already using it to whip up fake restaurant receipts. Yeah, you heard that right. It’s like giving fraudsters a new toy for their deepfake toolkit.

Deedy Das, that social media guru and VC bigwig, went ahead and posted a bogus receipt for a legit San Francisco steakhouse on X, all thanks to the 4o model. Talk about a wake-up call—’real images’ as proof? Not so reliable anymore, huh?

And it doesn’t stop there. Folks online are getting creative, adding ‘authentic’ touches like food stains to these fake receipts. Sure, some slip-ups (like math that doesn’t add up) give them away, but the sheer ease of creating these forgeries? That’s the real kicker. Fraudsters must be rubbing their hands together.

OpenAI’s response? They’re quick to point out that every image from ChatGPT comes with metadata shouting ‘I’m AI-made!’ Plus, they’re all about sticking to their guns on usage policies and tweaking things based on how people actually use (or misuse) the tech. They’re also tossing around ideas like using this for teaching financial literacy, or in art and ads. Because, you know, it’s not all doom and gloom.