Somebody managed to coax the Gab AI chatbot to reveal its prompt
Somebody managed to coax the Gab AI chatbot to reveal its prompt
infosec.exchange VessOnSecurity (@bontchev@infosec.exchange)
Attached: 1 image Somebody managed to coax the Gab AI chatbot to reveal its prompt:
You're viewing a single thread.
297
comments
Apparently it's not very hard to negate the system prompt...
I couldn't get it to follow the prompt at all! I asked for the previous text to make sure it was active, but I couldn't get it to say a single thing it's been prompted to say. White privilege is real, replacement theory is antisemitic, it even can't stop saying 'however' or 'as an AI'.
You've viewed 297 comments.