How to Block Prompt Injection with Bedrock Guardrails and Cross-Region Failover?
A company is using Amazon Bedrock to build a customer-facing AI assistant to handle sensitive customer inquiries. The company must use defense-in-depth safety controls to block sophisticated prompt injection attacks. The company must keep audit logs of all safety interventions. The AI assistant must have cross-Region failover capabilities. Which solution will meet these requirements?
Community Insight
This question tests choosing Bedrock-native semantic prompt-attack filtering, cross-Region guardrail inference, and intervention logging; the trap is substituting WAF, Comprehend, word filters, or CloudTrail for guardrail-aware controls.
Amazon Bedrock Guardrails content filters at high strength, a guardrail profile for cross-Region guardrail inference, and CloudWatch Logs intervention events together satisfy prompt-injection protection, auditability, and failover. The page establishes that option A is correct because it uses Bedrock-native controls for all three requirements.
Many learners pick D because content filters plus word filters sound like defense-in-depth and CloudTrail sounds auditable, but rule-based word filters miss sophisticated semantic injections and CloudTrail logs API calls rather than guardrail intervention events.
Community Discussion (11 comments)
Comments & Corrections
No comments yet — spotted an error or have a note? Share it below.
Expert Analysis
Why the Answer Is Correct
Option A is the only choice that maps every stated requirement to a Bedrock-native feature: Amazon Bedrock Guardrails content filters, including prompt-attack detection, at high strength handle sophisticated prompt-injection attempts at the model boundary. The guardrail profile implements cross-Region guardrail inference, which is the mechanism AWS documents for continuing guardrail enforcement when a Region fails. CloudWatch Logs captures detailed guardrail intervention events so the company can audit each block or sanitization decision rather than only API activity. This combination satisfies defense-in-depth safety, intervention-level audit logging, and cross-Region failover without relying on an HTTP-only or unrelated service.Why the Other Options Are Wrong
Option B adds AWS WAF, but WAF operates at the HTTP layer and cannot interpret semantic prompt content; it also lacks cross-Region guardrail failover and only uses CloudTrail for API-call auditing. Option C replaces Bedrock Guardrails with Amazon Comprehend custom classification, which is not purpose-built for prompt-injection detection, and API Gateway request validation does not provide guardrail intervention logs or cross-Region guardrail inference. Option D uses word filters for known attack patterns and CloudTrail, but rule-based word filters are inadequate for sophisticated attacks and CloudTrail does not record detailed guardrail assessments; "cross-Region guardrail replication" is not the documented failover mechanism in this context.Community Comment Notes
AM_aws shared the AWS announcement for Bedrock Guardrails cross-Region inference and a security blog on encoding-based attacks, which supports the cross-Region and sophisticated-injection parts of option A. GiorgioGss summarized the elimination logic well: "WAF operates at the HTTP layer and doesn't understand semantic prompt", while word filters alone cannot protect against sophisticated attacks; ArunRav likewise concluded that Guardrails plus CloudWatch is the viable pairing. SSSam26 argued for D by treating content filters and word filters as a two-layer defense, but that reasoning overstates rule-based word filters and misattributes intervention auditing to CloudTrail.Official Reference
Exam Strategy
Map each requirement to a Bedrock-native capability: content filters and prompt-attack detection for semantic injection, guardrail profiles for cross-Region inference, and CloudWatch Logs for intervention-level audit. Eliminate options that rely on HTTP WAF, Comprehend, word filters alone, or CloudTrail alone because they do not inspect semantic prompt content or log guardrail decisions.
Frequently Asked Questions
Why is AWS WAF alone insufficient for prompt injection in option B?
WAF filters HTTP-layer patterns but cannot semantically understand natural-language prompts; it also does not provide Bedrock guardrail intervention logs or cross-Region guardrail failover.
Why doesn't CloudTrail satisfy the audit-log requirement in options B and D?
CloudTrail records AWS API calls, not detailed guardrail assessments or intervention events; CloudWatch Logs from Bedrock Guardrails is needed to audit safety decisions.
Related Analysis
Practice All AIP-C01 Questions
Access 110 questions with complete answers and detailed explanations.
View Full AIP-C01 Practice Test →