AI0-001 AI Governance and Ethics Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Differential privacy applied to the combined dataset

  Why it's wrong here

  Differential privacy adds noise to query outputs or gradients, but applying it to a combined dataset still requires centralising every hospital's records, violating the no-sharing requirement. Federated learning keeps data local. This approach would be correct when a single trusted aggregator already holds the pooled data.

• ✗

  Centralising all data in one location and anonymising it

  Why it's wrong here

  Centralising and anonymising still requires moving raw records to one site, breaching the no-sharing requirement and creating a single breach target. Federated learning trains locally and exchanges only model updates. Central pooling would be correct when data-sharing agreements and governance permit consolidation.

• ✓

  Federated learning

  Why this is correct

  Federated learning trains a shared model across decentralised sites by exchanging model updates rather than raw records, so each hospital's patient data stays local. This satisfies the privacy constraint while still producing a diagnostic model from multi-hospital data.

• ✗

  Using pseudonymisation and then pooling the data

  Why it's wrong here

  Pseudonymisation replaces identifiers but the records still leave each hospital and are pooled centrally, so raw patient data is shared and remains re-identifiable. Federated learning keeps data local, sharing only model updates. Pooling would be correct when consolidation is contractually permitted and privacy risk is accepted.