Mastery card
I can explain Jailbreak
Tricks that try to make an AI ignore its safety rules.
A jailbreak is a prompting technique aimed at bypassing a model's safety policies to elicit disallowed outputs.
Sticky trick
Jailbreak = pick the lock on guardrails. Defences must evolve.
Take into the room
In plain English, what is Jailbreak?
Or open the share card 🔓
Or start your own 5-word trail from home.