Output
Free tool
PII Redactor
Strip personal data and secrets from a text before you paste it into ChatGPT, forward a ticket, or share a log. Emails, phones, IBANs, Belgian national & VAT numbers, card numbers, IPs, API keys — detected with checksums where they exist, replaced with tags, reversible pseudonyms or a mask. Everything runs in your browser; nothing is uploaded.
Input
Text to redact
Findings
What was redacted
Mapping
Pseudonym → original
Keep this table private — it is the key to reversing the redaction.
How it works
Redaction you can actually use
Checksum-validated identifiers
IBAN (mod-97), payment cards (Luhn), Belgian national register numbers and enterprise/VAT numbers are verified, so version numbers and random digits are left alone.
Reversible pseudonyms
EMAIL_1, IBAN_2… the same value always gets the same label, so the text stays coherent for a model or a colleague. You keep the mapping table; nobody else sees it.
Your own terms
Add names, companies or code names to redact them too — the categories a regex can never know about.
Secrets, not just PII
API keys, GitHub and AWS tokens, JWTs and private-key blocks are caught before they leak into a chat.
Built for Belgium & the EU
Rijksregisternummer / numéro national, BE VAT, IBAN and +32 phone formats work out of the box; everything else is generic.
Private by design
Static page, no server, no analytics on your text. Your category preferences are stored in your own browser only.
Questions
Is this GDPR-compliant anonymisation?
It is a redaction aid, not a legal guarantee. Regexes cannot find every name or address, and pseudonymised data is still personal data under GDPR while the mapping exists. Use it to reduce exposure, and review the output.
Why are names off by default?
The name detector is a heuristic (capitalized word pairs) and produces false positives on ordinary text. Turn it on when needed, or — better — list the names in "Extra terms".
Which replacement style should I use for AI prompts?
Pseudonyms. The model can still reason about EMAIL_1 vs EMAIL_2 or CUSTOM_1 (a client) vs CUSTOM_2 (a supplier), and you can map the answer back with the table.
Can it process PDFs or Word files?
Not in the browser version — plain text formats only (txt, md, csv, json, html, log). For documents at volume, see below.
Need this on thousands of documents, with real name/address recognition?
That is a project we do regularly: NER-based PII detection (multilingual FR/NL/EN), document pipelines, on-premise deployment. Talk to us.
Need redaction at scale?
Named-entity recognition for names and addresses in French, Dutch and English, PDF/DOCX pipelines, audit logs — deployed on your infrastructure.