A company is using Amazon Bedrock to build a customer-facing AI assistant that handles sensitive customer inquiries. The company must use defense-in-depth safety controls to block sophisticated…

Question

Options

• AConfigure Amazon Bedrock guardrails with content filters set to high to protect against prompt
• BConfigure Amazon Bedrock guardrails with content filters set to high. Use AWS WAF to block
• CDeploy Amazon Comprehend custom classifiers to detect prompt injection attacks. Use Amazon
• DConfigure Amazon Bedrock guardrails with custom content filters and word filters set to high.

• A
  75% (15)
• B
  5% (1)
• C
  15% (3)
• D
  5% (1)

Explanation

Amazon Bedrock Guardrails provides a multi-layered defense-in-depth approach to prompt injection: content filters, denied topics, word filters, and prompt attack detection all operate as configurable safety layers. Setting content filters to HIGH maximizes sensitivity for sophisticated injection attempts. Every guardrail intervention is logged to Amazon CloudWatch and optionally to Amazon S3, satisfying the audit log requirement. For cross-Region failover, Amazon Bedrock cross-Region inference profiles automatically route requests to available Regions when the primary Region is unavailable, fulfilling the failover requirement without custom routing logic. Option B adds AWS WAF, which operates at the HTTP/network layer and cannot inspect LLM prompt semantics - it does not add meaningful protection against prompt injection. Option C (Comprehend custom classifiers) requires training, maintenance, and custom integration, adding operational burden. Option D is similar to A but lacks the cross-Region failover mechanism.

No community discussion yet for this question.