AI Glossary
Guardrails
What is Guardrails?
Safety or policy constraints applied to AI systems to prevent harmful, misleading, or off-topic outputs. Can be implemented via system prompts, model fine-tuning, output filters, or external classifiers.
Example in practice
A customer service team deploying an AI assistant would add a guardrail in the system prompt preventing the model from discussing competitor pricing — keeping every response on-policy.
Learn more
See Guardrails applied in a professional context through this free course.
AI Strategy for Leaders →