Anonymizing Personally Identifiable Information

Overview

Parloa provides robust features for anonymizing sensitive data, ensuring the privacy and security of personally identifiable information (PII). This process includes the anonymization of data such as names, birthdates, driver's licenses, addresses, IDs, phone numbers, emails, bank numbers, and other sensitive information.

Parloa uses industry-standard protocols to identify and manage PII data.

The anonymization of PII is currently applicable only to the Training screen. The feature is enabled by default for all projects and can be turned off using the toggle on the the screen.

Anonymization Process

Parloa replaces actual PII with augmented data to protect your privacy. For example:

  • Names are replaced with different names.

  • Birthdates are replaced with different dates.

All sensitive data is similarly replaced with realistic but fictitious information.

Opt-in/Opt-out of the feature

There is now an option to opt-in or out of anonymizing PII training data at Project level must you choose it. This can be done easily with the toggle option on the training screen. Once the toggle is turned off, the PII data will no longer be anonymized and replaced with fake realistic data.

Previously anonymized values will not be unanonymized. Once the feature is toggled off, the change takes effect for the future unassigned utterances.

Highlighting anonymized values

The anonymized values are now highlighted with a different colour to bring transparency.

The values are highlighted provided the feature is enabled and only future PII Data will be highlighted once anonymised.

Past PII data, (even if anonymized) are not highlighted.

Data Retention

Training Screen

In the Training tab, when you click on one of the utterances in the table, the displayed utterance is anonymized to ensure privacy. When adding assigned utterances to an intent, the anonymized utterance will be assigned.

However, if you have permissions to Debug or Edit Project enabled for your role and you click on the utterance, the context of the conversation is displayed in a pane on the right. In this pane, the content remains as it was originally said, not anonymized.

To view the context sidebar of a conversation and access analytics data, the Debug permission must be enabled for the respective role.

FAQs

How would this PII feature be enabled?

This feature is enabled by default on the Training screen for all new unassigned utterances, and one can opt-out of this feature using the toggle. The feature can be turned off or on for a Project.

What happens to the PII data shown in the Debugger?

The handling of PII data in the Debugger remains unchanged. To restrict access to PII data on the Debugger, please provide the appropriate access permissions to your users.

What languages are supported to anonymize PII data?

Currently, English (US & UK) and German (DACH) are supported.

How long does it take to anonymize PII data in unassigned utterances, and could there be a delay in displaying these on the training screen?

The anonymization process for unassigned utterances can take up to 2 minutes. During this time, the utterances may not be displayed on the Training screen until the anonymization is complete.

Last updated