Hide personal data from an AI chatbot, then put it back in the answer
For the e-mail, contract or meeting notes you want a chatbot to summarise, translate or answer, without handing it the customer’s name and contact details. The same value always gets the same placeholder, so the chatbot can still follow who is who; you review every finding, and the mapping back to the real data stays in this tab unless you export it.
- Runs entirely in your browser
- Works offline
- No upload
- No sign-up
- No watermark
Loading the tool…
How it works
- Paste the text or drop a .txt or .md file. Add the terms that must always be hidden: people, customers, project names.
- Check the table of findings: switch off a row to leave a value visible, or add a word the detector missed. Read the scrubbed text, then copy it into the chatbot.
- Paste the chatbot’s answer into step 2: the placeholders are replaced with the original data. Export the mapping as a JSON file if you need it later, and clear it when you are done.
Why nothing is uploaded
Every operation on this page is done by code running inside your browser tab, using the same engine that renders web pages. The file is read from disk into your tab’s memory, transformed there, and written back out as a download. It is never sent anywhere — not to us, not to a third party.
Verify it yourself
- Open your browser’s developer tools (F12) and select the Network tab.
- Load your file and run the tool.
- The only requests you will see fetch the tool’s own code — and, for a few heavy tools, their open-source engine (also from this site) or an AI model from Hugging Face — plus one small page-view ping to loreatec.jp (page address and title, nothing more). None of them carry your file.
Frequently asked questions
Is the text anonymous after this?
No, and the tool does not claim it is. It is pseudonymisation: the data you can see is replaced with placeholders, but detection is automatic and misses things — a name written in an unusual way, a nickname, a detail in a sentence (“our only customer in Sapporo”) that still says who someone is. Put everything that must never appear in the term list: a term is found regardless of upper or lower case, full-width or half-width forms, accents and invisible characters, but not when it is spelt or abbreviated differently. Even so, read the whole scrubbed text before you paste it anywhere.
What does it find?
E-mail addresses; phone numbers in international and Japanese formats; web addresses and domain names; IPv4 and IPv6 addresses; payment card numbers that pass the Luhn check; IBANs with a valid checksum; Japanese My Number and corporate numbers with a valid check digit, and invoice registration numbers (T + 13 digits); Japanese postal codes and addresses; names of people, organisations and places in Japanese text, using a morphological dictionary; and, in Latin-script text, organisations by their legal form (Inc., Ltd, GmbH, S.L., B.V.…) and people after a title or in capitalised word sequences. The last two are a heuristic and are marked “guess” in the table.
How does it put the data back?
Each distinct value gets one placeholder, such as [COMPANY_2], and the tool remembers which is which. When you paste the answer, it replaces every placeholder with its original. Chatbots sometimes alter placeholders, so these forms are put back too: any mix of upper and lower case, spaces inside the brackets, a hyphen, a space or nothing instead of the underscore, leading zeros in the number, brackets escaped for Markdown, full-width characters, and the upper-case form without brackets (COMPANY_2) wherever it is not joined to a letter, digit or underscore, code included. A placeholder the chatbot translated or renamed ([PERSONA_1], [人物_1]), and any other word in square brackets followed by a number that is not in the mapping, is left as it is and listed. A chatbot can also put a placeholder in the wrong place, so read the restored text before you use it.
Where is the mapping kept?
In the memory of this browser tab only. It is gone when you close or reload the tab, or press Clear everything. If you need it later, export it as a JSON file and import it again; that file contains the original data, so treat it like the original text. The exported file holds the values of the text now in step 1 and of any mapping you imported, not values you only half typed or left visible. Importing a file replaces the mapping in this tab, after asking you. Nothing is stored in the browser unless you tick the option to remember your term list, and then only that list is stored.
Is my text uploaded?
No. Detection, the mapping and the restore all run inside this tab, and the page makes no request with your text. For Japanese text, a dictionary of about 18 MB is downloaded once from this site and kept by the browser.