ISACA Certified AI Security Manager — Question 4

A large pharmaceutical company using a new AI solution to develop treatment regimens is concerned about potential hallucinations with the introduction of real-world data. Which of the following is MOST likely to reduce this risk?

Answer options

• A. Penetration testing
• B. Data asset validation
• C. Human-in-the-loop
• D. AI impact analysis

Correct answer: C

Explanation

The 'Human-in-the-loop' approach is the most effective way to mitigate the risk of hallucinations because it involves human oversight in the AI decision-making process, allowing for validation and correction of outputs. The other options, such as penetration testing, data asset validation, and AI impact analysis, do not specifically address the issue of hallucinations in AI outputs related to real-world data.