Detecting Jailbreaks and Adversarial Inputs

This lesson is for enrolled learners

The concept Jailbreaks and prompt injections are attempts to override your instructions or to smuggle commands through untrusted text: the user message, a retrieved document, a web page, or a tool result. Detection is useful, but it is not the whole control. The robust pattern is to treat untrusted text as data, to enforce policy in code, and to use……

Enroll to continue Already enrolled? Sign in