Skip to content

AIP-C01 · Study Guide

← Domain 3: AI Safety, Security, and Governance

3.2.3 · Lesson 8 of 15

PII Protection

Detect, redact, or anonymize Personally Identifiable Information during model interactions using Guardrails and Comprehend.

WHAT

Detect, redact, or anonymize PII during model interactions.

PII handling modes

ModeBehaviorTypical use
ANONYMIZEReplace sensitive values with placeholders like [NAME], [EMAIL], [PHONE]Customer service workflows that must continue
BLOCKReject the entire request if any configured PII entity is detectedHard-stop for high-risk identifiers (e.g., credit cards)

HOW

Configure PII entity actions in Bedrock Guardrails and/or use Amazon Comprehend for pre-checks.

NAME / PHONE / SSN → ANONYMIZE
CREDIT_CARD_NUMBER → BLOCK

EXAM TRAP

If the requirement is to obscure email but allow the conversation to proceed, choose ANONYMIZE, not BLOCK.

Key points

  • PII ANONYMIZE vs BLOCK
  • Bedrock Guardrails PII actions
  • Amazon Comprehend entity detection
  • [NAME] / [EMAIL] placeholders
  • Continue vs terminate request

Exam focus

ANONYMIZE masks PII with placeholders and lets the conversation continue; BLOCK rejects the request outright when PII is detected.

View progress