The technical requirements for using trainable classifiers are any one of these licenses:

• Microsoft 365 E5 license
• Microsoft 365 E5 Compliance license

More information about the licensing requirements for using trainable classifiers can be found at the following link:

https://docs.microsoft.com/en-us/microsoft-365/compliance/classifier-get-started-with?view=o365-worldwide#licensing-requirements

When it comes to administrative privileges required to create and manage classifiers, you will need the following.

To access classifiers in the user interface, the global admin will need to perform an opt-in for the tenant in order to create custom classifiers.

Compliance administrator privileges are required to perform the training of a classifier.

You will need these permissions to use classifiers in the following scenarios:

• Retention label scenario:

a. Record management role

b. Retention management role

• Sensitivity label scenario:

a. Compliance data administrator

b. Compliance administrator

c. Security administrator

• Communication compliance scenario:

a. Insider risk management admin

b. Supervisory review administrator

These privilege requirements are listed at the following site:

https://docs.microsoft.com/en-us/microsoft-365/compliance/classifier-get-started-with?view=o365-worldwide#permissions

Now that we have covered the requirements part, let’s dive into our first topic and start exploring what trainable classifiers are.

What are trainable classifiers?

There are three ways to classify content in your Microsoft 365 tenant:

• Manually: The responsibility of classifying the data falls on the end user or the administrator, using the sensitivity labels created in the previous chapter.
• Automated pattern matching: Automated pattern matching includes finding content by using the following:

a. Metadata values or keywords

b. Sensitive information types to identify data

c. Document fingerprinting to recognize data

d. Exact data match by finding the presence of exact strings

• Trainable classifiers: The classification method of trainable classifiers is very well suited for content that is not easily identified by automated pattern matching or the manual approach.

Trainable classifiers are fed examples of data you want to train the classifier to identify, meaning that the classifier looks at data you provide in order to train its identification algorithm.

There are two separate types of classifiers available in Microsoft 365, the pretrained classifiers created by Microsoft, which you can start using directly without the need to train them, and the custom classifiers, which you create on your own if there are classification needs within your organization where the pretrained classifiers are insufficient.

We will explore each of these classifiers in the following sections.