AI 데이터 거버넌스 퀴즈

퀴즈를 풀면 정답과 설명이 나타납니다.

1. A dataset may be used for training and evaluation. What must be checked?
   • Whether its file opens in the training pipeline
   • Whether its permissions cover each intended use separately
   • Whether both uses require the same model
   • Whether the dataset has a large number of rows
2. Why does data governance track embeddings and cached responses?
   • They contain no source-related information
   • They are always public
   • They cannot be deleted
   • Derived assets may retain information or dependencies from source data
3. A deletion procedure removes the uploaded file but not its index copies. What remains unresolved?
   • The model’s user interface
   • The source file’s format
   • Deletion across relevant derived stores and copies
   • The data catalog title
4. Who should approve changes to data schema and validation rules?
   • An assigned owner responsible for quality and change control
   • Any user who uploads a file
   • The AI model itself
   • The storage vendor only
5. What does lineage provide?
   • Proof that data use is permitted
   • A guarantee of data quality
   • A substitute for retention rules
   • A way to trace where data and derived outputs came from
6. A dataset was collected for retrieval but is now proposed for model training. What is needed?
   • Assume access for one use implies every use
   • Review whether its permissions and purpose cover training
   • Delete the data immediately
   • Change the dataset label only
7. What should a correction procedure specify?
   • Who handles it, how it propagates, and how completion is verified
   • Only where the data catalog is stored
   • The model’s average token count
   • Whether an error appears in a presentation
8. Why revisit data access and retention when integrations change?
   • Integrations never affect data flow
   • Retention applies only before deployment
   • New services can change who sees the data and how long it persists
   • A new connector automatically narrows access