Syllaro Academy

AWS Certified AI Practitioner

Guardrails for Generative AI

How Amazon Bedrock Guardrails enforces safety policy in the request path: the six policy types, filter strengths, contextual grounding thresholds, and the ApplyGuardrail API.

Intermediate 22 minutes 5 Learning Objectives
  1. Explain why a guardrail enforces policy that a system prompt only requests
  2. Describe the six Amazon Bedrock Guardrails policy types and what each one catches
  3. Interpret content filter strength levels against confidence classifications
  4. Configure contextual grounding and relevance thresholds to block hallucinated responses
  5. Identify when to use the ApplyGuardrail API instead of attaching a guardrail to a model call