Security classification

Security classifications help you work safely with different types of information. With them you control which AI models and tools may be used in which context.

Why security classifications are needed

In Intric you create Spaces, and each Space can carry its own rules and boundaries — its security classification. Which Space you work in, and which rules apply there, depends on what the use case needs and requires.

Examples:

• A use case involving more sensitive information. Here the data has to stay within a given jurisdiction, so you need to limit which models and tools are used.
• A use case handling open, public information. Here it matters less where data goes, and you may instead want access to the newest models or a particular integration.

Security classifications let you serve both on the same platform.

How users notice it: they have access to different models and knowledge sources depending on which Space they are in, and they see a label in the chat reminding them what type of information applies there.

Enable security classification

You have to be an administrator.

1. Open the admin view
2. Click Security classification
3. Enable security classification

The feature is now on and you can start creating classifications.

Which classifications should you create?

That is up to you, but it is common to start with two:

1. Sweden or EU — where data flow is restricted to that jurisdiction
2. Open/Public — where more models and integrations may be used

You can change these and add more later.

Classifications are hierarchical

This is the most important thing to understand before you start:

• What is in the highest classification is also in the ones below it.
• What is in the lowest classification is not in the ones above it.

Example: if an AI model is in the Sweden classification, it is also in the Open classification. But models that exist only in Open do not appear in Sweden.

That is why it matters that you put the classifications in the right order.

Create a new security classification

1. Click Create new security classification
2. Fill in the label. It is visible to every user and reminds them what information may be used. Example: “Only open and public information may be used here”
3. Write instructions (optional). Describe what the classification means — this helps other administrators.

Configure models and tools

Once the classification exists, the next step is to attach models to it:

1. Click Configure completion models
2. Choose a security classification for each model — decide whether the model belongs to (in this example) Open or Sweden/EU
3. See where the model runs by clicking its name. You can also go to AI models in the left menu for a combined overview
4. Repeat for embedding models, transcription models and tools/integrations

Restrict a tool to ONE classification

Sometimes you want a tool used only in a specific classification and not carried down the hierarchy.

Do this:

1. Go to the left menu
2. Click AI tools
3. Click Edit on the tool
4. Scroll down
5. Enable Require matching classification

The tool is now used only in the chosen classification — not in the ones below, regardless of hierarchical order.

Use the classification in a Space

Once everything is configured, the classifications can be put to work:

1. Go into a Space (or create a new one)
2. Click Settings in the Space
3. Choose a security classification

The Space is now constrained: only the models, tools and integrations approved for that security classification can be used when someone builds assistants there.

Test your knowledge

1Question 1 of 32Question 2 of 33Question 3 of 3

Question 1 of 3

Question 1 of 3

Why do security classifications exist?

Why do security classifications exist?

• To make the interface look better
• To limit which models and tools may be used depending on how sensitive the information is
• To hide assistants
• Only for developers

Back Next