Microsoft Security Product Hub

Trainable Classifiers


Trainable Classifiers can be used where manual or automated keyword/metadata methods of classification do not work very well. Trainable Classifiers use machine learning to identify classified or sensitive information by evaluating hundreds of examples of previous documents and building a model. The classifiers can be run against content in Microsoft 365 and label content or apply retention labels based on the policy.


  • Automated classification of data that does not have easy keyword or sensitive information type matching
Secure Remote Work
Zero Trust