ISACA Certified Artificial Intelligence Auditor (CAIA) — Question 45

An AI tool is being implemented for a regional healthcare organization. Which of the following training methods BEST ensures the AI output does not reveal whether someone's personal data was used?

Answer options

• A. Supervised learning with labeled patient records
• B. Data augmentation during training to improve privacy
• C. Differential privacy applied during model training
• D. Transfer learning using public health data sets

Correct answer: C

Explanation

The correct answer is C, as differential privacy ensures that the output of the model does not disclose information about any individual in the training dataset, thus protecting personal data. The other options either involve using identifiable patient data directly or do not provide the same level of privacy assurance.