Jailbreak
An input crafted to make a model ignore its safety training or system instructions, for example through role-play, obfuscation or many examples, so that it produces content or actions it should refuse. It is a form of prompt injection; model refusals are one layer of defense, not a security boundary.
Covered in
Guide
Contain prompt injection at the application boundary
Assume retrieved content can influence model output; enforce data access and consequential actions in components the model cannot authorize for itself.
Guide
Build evaluations around prohibited outcomes
Test the application’s actual security boundaries, preserve legitimate tasks and distinguish deterministic enforcement from variable model behavior.
Practice: Local project: Gate model and provider changes on a versioned red-team corpus