Amazon Bedrock Guardrails helps apply configurable safeguards to generative AI applications. For AWS exam questions, recognize it when a team needs to reduce harmful or inappropriate responses, detect sensitive information, or apply topic restrictions to model inputs and outputs.
What Guardrails can help control
- Denied topics: define subjects an application should not discuss.
- Content filters: detect and filter supported harmful categories of text or image content.
- Sensitive information: detect, mask, or block supported personally identifiable information and other configured sensitive data.
- Word filters: block configured words or phrases.
- Contextual grounding checks: help assess whether responses are grounded in provided sources and relevant to the query.
Exam decision clues
If a scenario asks for configurable safeguards around prompts and model responses in an Amazon Bedrock application, consider Guardrails. If the requirement is to explain why a model is biased or evaluate model behavior, related services such as SageMaker Clarify may be more relevant. Guardrails reduces risk; it does not guarantee that every response is correct or safe.
Related topics
Continue with foundation-model hallucinations, SageMaker Clarify, and Amazon Q Business vs. Amazon Q Developer.
Practice tip: identify whether the requirement is to filter content, protect sensitive data, evaluate model quality, or ground answers in a knowledge source before choosing a service.
Prepare for the AWS Certified AI Practitioner (AIF-C01)
Test your understanding of AWS AI services and responsible AI concepts with original practice questions and detailed explanations.
Try the free AIF-C01 diagnostic or explore the 390-question premium practice package.