What is guardrails?

Rules, checks, or limits designed to reduce harmful or unwanted AI behavior. They help, but can fail.

Also called Guardrail, AI safeguards.

For example: A guardrail blocks an assistant from making a purchase without approval.