Keeping people away from data: A generative AI use case
summary
Argues that many projects should delete data when not needed rather than store it; when storage is required the recommended pattern uses SageMaker RAG + OpenSearch vectors + S3 knowledge base and protects data with identity controls (IAM tags/groups, IAM Identity Center) and data controls (KMS encryption, ACM/Private CA for transit, Macie detection, Comprehend for PII redaction via EventBridge+Lambda). Emphasizes keep PII out of RAG, use defense in depth, and shows trusted identity propagation feeding CloudTrail so audits trace a specific user across EMR, Glue and S3.
AI-written summary · about
chapters (5)
- 0:00Introduction and architecture
- 4:37Identity-based access
- 9:39Data protection strategies
- 13:03Auditing and traceability
- 17:30Summary of best practices
related talks
25mSafeguarding Sensitive Data in the Age of Generative AI
40mSecuring Generative AI: Is it all an Illusion?
26mGenerative AI driven sensitive data protection
24mStop Making Your Data Team the 'Data Police'
20mWhy data privacy is critical for robust data management?
30mStreamlining data governance
42mRise of the Machines: Achieving Data Security and Analytics with AI
09mSecuring Generative AI Workloads