quiz bu lëkkale ak tegtal · Dëgër Niveau
IA & Laajteb Njaboot
Dafay saytu ni done yi di jaar, wuute ci wàllu nëbbëtu, jàpp ni bokk na ci kuréel gi, ràññeewaat ci anam wu jaarul yoon, efaase ak saytu risku nëbbëtu.
Laaj 1 de 8
Which statement best distinguishes security from privacy?
Xoolaatal laaj bu nekk8 laaj
Sooy def quiz bi, tontu yi ak leeral yi dañuy feeñ.
- Which statement best distinguishes security from privacy?
- Security and privacy are identical concepts
- Encryption alone establishes consent
- Secure storage can still be a privacy problem if data are excessive or used unexpectedly
- Privacy only concerns passwords
- In a differential privacy analysis, what does epsilon influence?
- The number of records in the input
- The privacy-accuracy tradeoff, with smaller epsilon generally giving stronger privacy
- The model’s training temperature
- The encryption key length
- A membership-inference attacker asks whether a person’s record was in training. What is being tested?
- Whether a user can reset a password
- Whether a record is duplicated in a database
- Whether a prompt contains a name
- Whether the model output leaks information about training-set membership
- Why can removing names fail to make a dataset anonymous?
- Other attributes or combinations can re-identify people
- A name is the only possible identifier
- Data cannot be linked across tables
- Removing names always satisfies every privacy goal
- Which data practice best follows the guide’s privacy principle?
- Keep every prompt indefinitely for any future use
- Collect only information needed for the stated task and set a retention period
- Copy all source records into each model call
- Treat local processing as proof of no privacy risk
- Why should a team inspect embeddings and indexes after deleting an uploaded document?
- Deleting a document automatically removes every copy everywhere
- Embeddings contain no data relationships
- Derived artifacts may still reveal information from the original data
- An index is only a user interface
- What should a team document about AI data processing?
- Only the model’s public name
- The number of support tickets
- A generic privacy slogan
- Which providers and internal services handle each data category
- Why does an assistant’s unsupported claim that data were deleted not suffice?
- The team should test deletion and account isolation through the actual application
- A chatbot response is a deletion receipt
- Data deletion is never relevant
- The assistant can inspect all vendor systems