Skip to content

Anonymization Preview

The Preview tab lets you paste any text and see exactly what your active policy redacts before a single real request touches the LLM. No vault record is created, nothing is logged.

How to use it

  1. Go to Preview in your Redact workspace sidebar
  2. Select a policy from the dropdown, or leave it on Default policy
  3. Paste or type your sample text
  4. Click Run Preview

Results appear immediately below.

What you see

Stats bar

A row of cards showing how many entities were detected, broken down by type, and which policy was applied.

Card Meaning
Entities found Total count across all types
Person, Email, Phone, etc. Count per entity type that fired
Policy Name of the policy that was used

Side by side

Original on the left, Anonymized on the right.

  • Detected PII in the original is highlighted so you can see exactly what the engine picked up
  • Pseudonyms in the anonymized output are highlighted so you can trace each replacement at a glance

Entity table

A full mapping of every real value and the pseudonym it was replaced with. Useful for verifying that names, emails, and custom patterns resolved to the right replacements.

Note on clipboard and the Chrome extension

If you have the Xybern Redact Chrome Extension installed with the clipboard scanner enabled, pasting text into the Preview textarea will trigger the scanner first. The extension detects PII locally, warns you, and anonymizes the text before it reaches the page. This means the Preview will receive already-anonymized text and run a second pass on it.

This is the expected behaviour of the clipboard scanner - it fires on every paste on every page, including the Preview tab. To test your policy against raw original text, type the sample directly into the textarea instead of pasting.

What the clipboard scanner actually does

The clipboard scanner is the most immediate layer of protection in the Xybern stack. The moment you paste text containing PII anywhere on any webpage, it intercepts before the text reaches the input field:

  1. You copy text containing sensitive data
  2. You paste it anywhere, any website, any form, any LLM interface
  3. The extension scans locally, no network request at this step
  4. A warning banner appears if PII is detected
  5. You click anonymize, only then is the text sent to Xybern
  6. The anonymized version replaces what you pasted

The sensitive data never reaches the page in its original form. For teams using LLM tools directly in the browser, this covers every workflow without requiring any changes to how they work.

What does not get logged

The Preview endpoint runs the full anonymization engine against your policy but skips vault storage entirely. No record appears in the Provenance Vault, no chain hash is computed, and no HMAC is signed. Use Preview freely for testing without polluting your audit log.