How to remove personal data before using AI
- Paste the text you want to share with ChatGPT, Claude, Gemini or a colleague: an email thread, a support ticket, a document or a log file.
- Email addresses, card numbers (checked with the Luhn algorithm), IBANs, phone numbers, US Social Security numbers, UK National Insurance numbers, IP addresses and postcodes are found and replaced. Tick links and dates too if needed.
- Add names, addresses, company or project names to the "also redact" box, since names in free text cannot be detected reliably by patterns.
- Copy the redacted text. If you used numbered tags, paste the AI's reply into "Restore" to swap the real values back in.
Why redact before prompting
Text you paste into an online AI service is sent to that company's servers, may be stored, may be reviewed by staff for safety, and on some consumer plans may be used to train future models. For personal data about customers, patients, employees or yourself, that can break privacy laws such as GDPR, HIPAA or your company's policy. Redacting first keeps the useful content, so the model can still summarise the complaint or draft the reply, while the identifying details never leave your device.
Numbered tags such as [EMAIL_1] keep the meaning of the text: the model can see that the same person appears twice, and you can put the real values back into its answer afterwards. Neutral values make text read naturally, and masks are best for sharing with people. Pattern-based detection is fast and private but not perfect: it can miss unusual formats and cannot recognise every name or address, so always read the redacted text before sending it. Everything, including the mapping, stays in this page and is lost when you close it.
Does this detect names automatically?
No. Reliable name detection needs a language model, which would mean sending your text somewhere. Add names to the list instead.
Is it safe to use with confidential data?
The page runs entirely in your browser and makes no network requests with your text. Follow your organisation's data policy too.
Can I undo the redaction?
Yes, with numbered tags: the restore box swaps the tags back for the original values while this page is open.