ISACA Certified Artificial Intelligence Auditor (CAIA) — Question 34

Which of the following is the MOST significant risk associated with a deep learning system algorithm being updated as it learns?

Answer options

• A. System algorithms can easily be modified by attackers because the algorithms reside in system memory.
• B. Project stakeholders may not endorse the system because its behavior may be contrary to their expectations.
• C. Operational risk may increase because the system is continuously running.
• D. The system may generate discriminatory output because of biases in training data.

Correct answer: D

Explanation

The correct answer is D because biases in training data can lead to discriminatory outputs, which is a significant ethical and operational risk. Option A is incorrect as it focuses on security concerns, while B relates to stakeholder perceptions, and C addresses operational continuity, none of which are as critical as the risk of bias in deep learning outputs.