Anyone interested can check the full HTML source here. We're pretty sure that the HTML we prompted ChatGPT with does not contain a malicious script: rather, it looks like the attributes of an ivory/champagne coloured (and possibly Japanese) dress for sale on an e-commerce website. The forbidden tokens seem to make GPT a little bit paranoid.
I think that here ChatGPT was actually (mostly) correct -- the code above this HTML is indeed malicious. I have manually cleaned up it myself, it's on pastebin here.
It is just a simple (but quite obfuscated) script,... (read more)
I think that here ChatGPT was actually (mostly) correct -- the code above this HTML is indeed malicious. I have manually cleaned up it myself, it's on pastebin here.
It is just a simple (but quite obfuscated) script,... (read more)