A control or safeguard placed around AI systems to limit harmful or undesired outputs.
"The platform added AI guardrails to block inappropriate language in customer chats."
