ISACA Advanced in AI Security Management Practice Mode

Community Discussion

Written by other learners. Comments can disagree with the marked answer — check the official documentation when in doubt.

• Chose C1

  Data bias in ML primarily arises from unrepresentative, incomplete, or skewed training datasets (e.g., underrepresenting certain groups, demographics, or scenarios), leading to unfair or inaccurate model outputs. The most direct and effective way to mitigate this is by diversifying the training data — collecting and using representative, inclusive, and varied sources to ensure balanced coverage across relevant populations, features, and conditions. This is a foundational pre-processing step emphasized in responsible AI practices and aligns with ISACA's AI governance guidance on addressing bias through high-quality, diverse data.