Anonymize Text Before Sharing It With AI
Replace names, emails, and account numbers with placeholders before the text goes to a chatbot - then restore them in the answer.
The problem with pasting real data into chatbots
"Summarize this client email." "Draft a reply to this complaint." The fastest way to use AI at work is to paste real correspondence into it - together with real names, phone numbers, addresses, and account details. Under Swiss data protection law (nDSG) and the GDPR, that can already be a disclosure to a third party, and many company policies forbid it outright.
The practical middle ground is pseudonymization: swap the personal data for placeholders before the text leaves your hands, and swap the real values back into the AI’s answer afterwards. The ASD123.ai Anonymizer automates exactly this round trip - locally in your browser.
How the round trip works
- Paste your text (or upload a .txt, .docx, or .pdf file). Detection runs instantly in your browser.
- Review what was found: names, emails, phone numbers, addresses, IBANs, credit cards, IDs, URLs, IP addresses, companies, dates, and more - grouped by category, with a highlight view and per-entity toggles.
- Copy the anonymized text, where every entity is a placeholder like [PERSON_1] or [EMAIL_1].
- Use it in ChatGPT, Claude, or any other service.
- Paste the AI’s answer back and click deanonymize - the placeholders turn back into the original values.
The mapping between placeholders and real values stays in your browser’s memory only - it is never written to disk or sent anywhere, and you can export it as CSV if you need to continue later.
Detection modes
Fast regex patterns cover structured data (emails, phone numbers, IBANs, credit cards, IDs) and common name patterns out of the box. For unstructured text you can optionally load a local AI detection model via WebGPU - the model files are downloaded once, but your text still never leaves the browser. Whichever mode you use, review the result before sharing: automated detection is an aid, not a guarantee.
Frequently asked questions
What kinds of personal data does it detect?
Names and titles, email addresses, phone numbers, street addresses, dates of birth, IBANs, credit card numbers, social security and ID numbers, companies, URLs, IP addresses, UUIDs, and other identifier patterns - each in its own category that you can switch on or off.
Is the detection really local?
Yes. Regex detection is plain JavaScript in your browser. The optional AI mode downloads model weights once from Hugging Face, but the inference runs locally - your text is never uploaded.
Can I restore the original names in the AI’s answer?
Yes. As long as the answer still contains the placeholders, one click replaces them with the original values. The mapping lives only in your browser session, and you can export it as CSV to continue later.
Is this enough for GDPR or nDSG compliance?
It is a strong technical aid, not a legal guarantee. Automated detection can miss context-dependent personal data (for example, "our CFO" identifies a person without naming them), so review the anonymized text before sharing anything sensitive.
Protect the data, keep the workflow
Free, reversible, and everything stays in your browser.
Related: Clean ChatGPT text · Run AI chat locally