3.2.2 · Lesson 7 of 15
Privacy-Preserving Systems
Design GenAI data paths that minimize exposure of sensitive information through discovery, retention controls, and privacy-preserving processing.
WHAT
Build privacy-preserving systems so sensitive data is discovered, governed, and expired appropriately across GenAI pipelines.
HOW
- Use Amazon Macie to scan source S3 buckets for sensitive PII prior to vector indexing.
- Enforce automated retention via S3 Lifecycle rules.
- Prefer redaction/anonymization before embedding when possible.
WHEN TO USE
- RAG corpora that may contain customer PII
- Regulatory retention limits on transcripts or chats
- Pre-index discovery of sensitive fields
Key points
- Amazon Macie for S3 PII discovery
- S3 Lifecycle retention rules
- Minimize sensitive data before indexing
- Privacy-preserving RAG pipelines
- Governance before vectorization
Exam focus
Use Amazon Macie to scan source S3 buckets for sensitive PII before vector indexing, and enforce retention with S3 Lifecycle rules.