Guardrails

Generative AI guardrails refer to a variety of different safety measures used to secure AI systems. Most commonly, guardrails are separate LLMs trained to detect malicious inputs or outputs, but guardrails can also be as simple as a keyword blacklist or an improved system prompt.