gallanon
Pseudonymize a document before sending it to AI
You already have a document with real details and want to work on it with AI. Load a TXT, MD, RTF, DOCX, ODT or text PDF, review the selections, then download a document with labels such as [PSN-1] and a key file. Editable files keep their format; a PDF becomes TXT or DOCX without the page layout. Later, load the AI response and your key to restore the real details.
How it works, step by step
- Wait for offline readiness. Prepare the optional PDF support if needed, then disconnect from the internet if you wish.
- Load your document. gallanon suggests passages to review without sending the text to an AI model.
- Read the preview. Accept or reject suggestions and mark missed details, including events, project names, amounts or contextual clues.
- Confirm your review and download the document and key. Password encryption is enabled by default. Store the key separately.
- Send only the reviewed, labelled document to your chosen LLM. Ask it to keep the labels. Later, load the AI response and key into gallanon, review the matches and download the restored text.
Occurrences resolved to the same entity share one label. Review whether full names, surnames and recorded variants belong together. Two people with the same surname may need manual separation.
TXT and MD are easier to inspect because they lack the layers of office documents. Supported DOCX, ODT and RTF formatting is retained. A text PDF produces TXT or DOCX: its text is extracted, but its visual layout is not preserved.
Example use
You want AI to review a draft tenancy agreement. Replace the parties, property address and numbers with labels. With the saved key, you can restore the real details in the revised agreement.
What to check
- Read the whole document. A rare diagnosis, specific transaction or unusual situation can identify someone without a name.
- Scans have no text for the detector. Use manual redaction for them. gallanon does not perform OCR.
- Some labels may be absent from a shorter AI response. Modified labels need confirmation; unknown labels need manual mapping.
- Pseudonymization is reversible and the data remains personal data. It does not establish permission to use a particular AI service.
Pseudonymized data remains personal data. A redacted document may also identify someone through its remaining context.