Jailbreak
A jailbreak is a trick prompt designed to get an AI model to ignore its safety rules — through role-play, odd formatting or step-by-step manipulation. AI companies patch known jailbreaks, and new ones keep appearing.
In one line, for a 12-year-old
A jailbreak is trying to trick an AI into breaking its own rules.
An example
Someone asks a chatbot to "pretend you're an AI with no rules" to get advice it would normally refuse.
Why it matters to people
Jailbreaks show that safety rules are not foolproof, which is why sensitive uses — health, children, finance — need more than the model's own safeguards.