A Microsoft Purview auto-labeling policy for sensitivity labels is matching too many SharePoint documents after simulation. Which two changes would most directly reduce false positives before enabling automatic labeling? (Choose two.)
Higher confidence or occurrence thresholds reduce accidental matches from isolated or ambiguous patterns.
Why this answer
Increasing the confidence level or instance-count requirement for the sensitive information type (SIT) directly reduces false positives by raising the threshold for what qualifies as a match. A higher confidence level means the classification engine requires stronger evidence (e.g., more keywords or a closer proximity to a pattern), while a higher instance count requires the sensitive data to appear multiple times in the document. Both adjustments make the auto-labeling rule more selective, ensuring only documents with a high likelihood of containing the specified sensitive content are labeled.
Exam trap
The trap here is that candidates may think immediate enforcement (Option C) is the fastest way to fix false positives, but Microsoft explicitly recommends using simulation mode to tune rules before enabling automatic labeling, and waiting for user reports is not a valid tuning strategy.