Currently available in public preview (MC348214), we’re introducing a compliant feedback mechanism to enable admins to report false positives for sensitive information types (SITs) and trainable classifiers detected and reported in Content explorer and DLP Alerts (non-legacy alerts). This Feedback Loop will improve the overall accuracy of SITs and trainable classifiers as it enables Microsoft to make systematic improvements in the product to reduce future false positives.
This message is associated with Microsoft 365 Roadmap ID 93210.
Now available in public preview, admins can provide feedback for sensitive information type (SIT) False Positives via Content explorer and DLP Alerts. This feedback can be provided in a compliant manner and users can see a summary view of the feedback in the SIT overview page. The feedback will be used to make systematic fixes to SITs for improved accuracy.
When this will happen
Rollout will begin in late November and is expected to be complete by end of December.
How this will affect your organization
It can be difficult to assess the accuracy of classifiers when there are false positives, and reporting false positives often involves sharing examples offline which is inconvenient and could risk exposing sensitive information. To address this challenge, we’re introducing Feedback Loop to seamlessly allow compliance admins to share feedback on policy mismatches in a compliant manner. Now in preview, this feature will soon be generally available for Content explorer and for DLP Alerts (non-legacy alerts).
Feedback Loop experience in Content explorer
In the Microsoft Purview compliance portal, admins can go to Content explorer > Contextual Summary to easily review classifier matches for a given file and mark them as true or false positives. Note: The contextual summary displays a limited set of matches and may not reflect the complete set of matches found on the item.
For items that are not a match, admins can opt to submit a redacted copy of the falsely matched item to Microsoft with feedback to improve classifier accuracy. Once submitted, there is also an option to withdraw the feedback.
Note: The new feedback experience is supported only for newly created or updated items. For items classified prior to this feature getting enabled for the tenant, the contextual summary and the feedback experience may not be available.
On the SIT overview page for a given classifier, admins can also provide feedback and see a summary of feedback submitted on the Feedback results tab.
What you need to do to prepare
No action is needed to enable this feature.
Get started with sensitive information types in the Microsoft Purview compliance portal.
Message ID: MC476041
Product(s): Microsoft Purview compliance portal
Cloud instance(s): Worldwide (Standard Multi-Tenant)