ISACA Certified AI Security Manager — Question 18

After deployment, an AI model's output begins to drift outside of the expected range. Which of the following is the development team's BEST course of action?

Answer options

• A. Return to an earlier phase in the AI life cycle.
• B. Take the AI model offline.
• C. Adjust the hyperparameters of the AI model.
• D. Create an emergency change request to correct the issue.

Correct answer: B

Explanation

The best action is to take the AI model offline to prevent further erroneous outputs while the issue is investigated. Adjusting hyperparameters or creating a change request may not address the immediate risk of incorrect outputs, and returning to an earlier phase could delay resolution.