Guardrail
Observation reveals behavior; guardrails intervene in selected behavior.
A guardrail is a check or control applied around an AI system’s inputs, outputs, or actions. It can block, transform, or route behavior according to a rule or a model-based assessment.
[LangChain]In practice
A support assistant masks email addresses in a draft before displaying it in a shared review queue.
[LangChain]A little deeper
Rule-based checks are predictable but can miss nuances. Model-based checks can handle more context but can also make errors. The term describes a family of controls, not a guarantee. [LangChain]
A common mix-up
A guardrail makes any agent action safe.
A check covers the cases it detects; other controls and evaluation remain necessary. [LangChain]
Helpful to know: Agent