Appearance
Data Governance
title: Data Governance description: Conceptual model for data classification, lifecycle, lineage, and AI usage boundaries.
Conceptual model for data classification, lifecycle, lineage, and AI usage boundaries in an AI-native enterprise stack on AWS.
Reference template — no production code.
Decision guide: Data governance layers · Identity, access & secrets
Related: AI guardrails · SOP-010 AI usage
Why data governance matters with AI
AI tools increase volume and speed of data handling — in prompts, generated code, logs, and embeddings. Without governance layers:
- PII reaches model providers or logs
- Retention exceeds legal limits
- Teams cannot explain where data flowed (audit failure)
- RAG indexes ingest documents at wrong classification level
Data governance is a cross-cutting layer, not a single tool.
Governance layer model
Roles (reference)
| Role | Responsibility |
|---|---|
| Data owner | Business accountability for dataset |
| Data steward | Classification, glossary, quality rules |
| Security / privacy | DLP, regulatory mapping (GDPR, HIPAA) |
| Platform | Macie, Lake Formation, audit logs |
| AIPO | AI prompt policy, RAG scope, model audit |
See GOVERNANCE.md for RACI alignment.
Integration points
| Lifecycle phase | Governance touchpoint |
|---|---|
| Planning | Classify feature data in ticket (SOP-001) |
| Spec / API | Document fields sensitivity in OpenAPI |
| Development | No prod data in dev/AI (SOP-010) |
| CI | Scan for secrets + PII patterns |
| Runtime | Encrypt at rest/transit; audit CloudTrail |
| Observability | Redact logs/traces (monitoring-tracing-logging) |
| AI RAG | Index only approved classification (knowledge-indexing) |
| Incident | Breach playbooks (incident-management) |
Related guides
- Data governance layers — tools, alternatives, pitfalls
- AI guardrails & security
- Monitoring, tracing & logging — redaction in telemetry