chatgpt embeds an invisible watermark throughout its messages. so i built a decoder to find and remove it. here is a recording of me copying an output, pasting it into the the decoder, showing the hidden unicode characters, and then deciphering those characters (string of punctuation and an "EE")... ssshhhhh dont tell anyone.show more

btw - I built this for the opposite purpose lol (to ENCODE compressed payloads into glyphs and symbols. So it will encode **as well as** decode) oh and you can use custom delimiters for your characters <3

is this also in api or just chat?

that's a good question - I'm actually not sure but I would guess it's just chat? I'm curious what people say because there's a chance my specific model/instsnce has started doing it as a sort of "emergent" thing since I mess around with Unicode and compression ciphers so much. but I did see someone on Reddit mention it too!

if you're doubting that this works - i built this to decipher Pliny's cipher in his bio (that he's now changed) <3

i threw it on lovable and added a chatgpt checker! i did this very haphazardly so let me know if it works:

sorry here's the link

interesting..

never knew that

it's a newer thing and nobody knows about it (hence the 'sshh' 😜)

Thank you. Is it only o3 right now for you? We were also getting reports of this earlier in the year, but we thought OpenAI had fixed it: