Anonymization

Administrators Only

This setting is only visible to users with the Administrator role.

The Anonymization settings let you configure exceptions to the automatic data masking in anonymous mode. By default, anonymous mode replaces all recognized personal data with placeholders before it is sent to the language model. Here you can exempt individual data types (e.g., names or emails) from anonymization — either entirely or only for specific values you define. Everything else remains masked.

Anonymous mode does not replace your own diligence

Anonymous mode is an additional protective measure and detects many common patterns such as names, addresses, or phone numbers – but not always all sensitive information. Remind your users to always review their inputs for sensitive content themselves. Anonymous mode replaces neither your users’ own duty of care nor your organization’s data protection policies.

Opening Anonymization Settings

1. Navigate to Admin Settings → Anonymization
2. You will see the Anonymization exceptions configuration panel listing all data categories

How Exceptions Work

Each row represents a category of personal data (e.g., person names, email addresses, organizations). For each category, you can:

• Enable the exception — data of this type will no longer be masked in anonymous mode
• Optionally restrict by pattern — only values matching a specific pattern are exempt; all other values in the category are still masked

Each category shows a placeholder and a plain-language example to help you understand what kind of pattern works for that data type. For instance, the email category suggests a pattern that exempts all addresses from a specific domain, while the person names category shows how to exempt a single name. You can use the | character to list several values in a single pattern.

When a category is exempt, users in anonymous mode will see that data passed through to the AI without masking.

Available Categories

Category	Description
Person names	First and last names of natural persons
Organizations	Names of companies, institutions, and public bodies
Locations & addresses	Cities, countries, streets, and addresses
Email addresses	Personal and organizational email addresses
Phone numbers	Landline and mobile numbers
URLs & IP addresses	Web addresses and IP addresses
Dates & times	Specific dates and times, e.g., dates of birth
Financial accounts	Credit card numbers, IBANs, and bank account numbers
Government IDs	Social security, passport, driver’s license numbers
Nationality, religion & politics	References to nationality, religious, or political affiliation
Other / unrecognized	Personal data that does not fit a known category

Using Patterns

A pattern is a kind of search template that lets you exempt only specific values within a category. Leave the field empty to exempt the entire category.

Two rules always apply:

• The entire value must match – the pattern has to cover the complete detected value, not just a part of it.
• Upper and lower case don’t matter – Miller and miller are treated the same.

The Most Important Building Blocks

Plain text matches itself exactly. In addition, a few special characters give you more flexibility:

Building block	Meaning	Example	Matches e.g.
Plain text	Exactly this text	Your City Administration	only “Your City Administration”
.*	Any number of any characters (including none)	.*@your-city\.gov	any address ending in “@your-city.gov”
.	A single arbitrary character	M.ller	”Miller”, “Moller”, “Muller”
\.	An actual dot	info\.dept	”info.dept”
\d	A single digit (0–9)	Ref-\d\d\d	”Ref-” followed by three digits, e.g. “Ref-123”
|	”or” – one of several values	Springfield|Shelbyville	”Springfield” or “Shelbyville”
( )	Parentheses group a part together	.*@(dept|city)\.gov	addresses ending in “@dept.gov” or “@city.gov”

Important: the dot

The dot . stands for any character. When you mean an actual dot (e.g., in a domain or file extension), write \. with a backslash in front of it. The same applies to other special characters such as *, |, (, or ): a preceding \ makes them be treated as a normal character.

Examples

• Exempt all email addresses from your authority: .*@your-city\.gov – .* stands for the arbitrary part before the @, followed by your fixed domain.
• Exempt multiple domains: .*@(your-city|other-town)\.gov
• Exempt a fixed organization name: Your City Administration – exactly this text, with no special characters.

Have Ayunis Core create the pattern for you

You don’t have to write regular expressions yourself. Simply describe in chat which values should be exempt and have the matching pattern created for you – for example: “Create a regular expression that matches all email addresses ending in @your-city.gov.” Then copy the result into the pattern field.

Pattern Validation

Patterns are validated when you save. If a pattern is not a valid regular expression or could cause excessive processing time, it will be rejected. The pattern must also not exceed the maximum character length.

Saving Changes

1. Enable or disable exceptions for the desired categories
2. Optionally enter a pattern to restrict the exception
3. Click Save

Changes take effect immediately for all users in your organization when they use anonymous mode.

Next Steps

• Anonymous Mode — How users work with anonymous mode in chat
• Security — Configure IP-based access restrictions
• Instructions — Set organization-wide AI instructions