Jailbreak
A jailbreak is a prompt crafted to make a model bypass its own safety training and produce output it would normally refuse.
Jailbreaks and prompt injection are often confused. A jailbreak targets the model’s training; injection targets the application’s instructions. Defences differ accordingly, and application-level prompt wording does little against either on its own.
Related terms
- Prompt injection
- Prompt injection is an attack in which instructions hidden inside content a model processes cause it to ignore its original instructions.
- Guardrail
- A guardrail is a check placed around a model — before input reaches it or after output leaves it — that enforces a rule the prompt alone cannot guarantee.
More on safety
See the full glossary, read the guides, or put it into practice in the prompt builder.