A compliance officer needs to demonstrate to regulators that the organization's AI training data for a credit scoring model does not contain prohibited personal attributes (race, gender, religion) that could lead to discriminatory model outcomes. Which Microsoft Purview capability can identify these attributes in the training data?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Here's the deal — a is correct because Microsoft Purview Data Map includes built-in sensitive information type classifiers that can detect personal attributes such as race, gender, age, and religion indicators within structured and unstructured data. Running a scan against the AI training data lake using these classifiers identifies where prohibited attributes exist in the training data, enabling remediation before model training.
Full explanation below image
Full Explanation
A is correct because Microsoft Purview Data Map includes built-in sensitive information type classifiers that can detect personal attributes such as race, gender, age, and religion indicators within structured and unstructured data. Running a scan against the AI training data lake using these classifiers identifies where prohibited attributes exist in the training data, enabling remediation before model training. B is incorrect because manual classification using custom tags requires human review of all data, which is impractical for large training datasets and does not provide automated detection. C is incorrect because sensitivity labels applied manually by data stewards require the same manual review approach, which cannot scale to identify specific attributes within large training datasets. D is incorrect because Compliance Manager fairness assessments document controls and policies around AI fairness but do not scan actual data to identify prohibited attributes.