AI & Data Governance
AI & Data Governance - what the exam tests
This domain covers Informatica's governance stack - business glossary, policies, stewardship, privacy classification and masking - plus how CLAIRE-based AI features (metadata-driven recommendations and generative assistance) support them. Expect questions on classifying sensitive data, applying policy outcomes, and governing AI and analytics use of data.
Core rules
- Governance artifacts: business glossary terms, data domains, policies, rules, and stewardship assignments. Glossary terms define meaning; policies state obligations; rules are testable and can link to data quality scorecards.
- Privacy flow: discovery scans sources for sensitive data, classification assigns categories, and policies trigger protection. Persistent masking changes stored values; dynamic masking rewrites values at query time; tokenization replaces values while preserving referential integrity for joins.
- Lineage connects policies and glossary terms to physical assets, enabling impact analysis before a schema change.
- Responsible AI: catalog training datasets, record model metadata, monitor input quality, and apply access policies to feature stores.
Worked example
A Snowflake table holds EMAIL, SSN, and DOB. Discovery classifies EMAIL and SSN as PII and DOB as sensitive personal data. The policy for non-production consumers requires dynamic masking, so analysts querying through the governed connection see masked email. A tokenized key column still joins correctly because tokenization preserves equality. Reasoning: the masking choice depends on whether consumers must join on the value - tokenization preserves joins, redaction does not.
Traps
- Confusing a glossary term with a policy: terms define meaning, policies drive enforcement.
- Assuming dynamic masking rewrites stored data; only persistent masking does.
- Hand-writing mask rules without classification, so newly added columns leak unmasked.