← Back to the tool

Don’t take our word for it — check

Jasyr processes the document in your browser. Below are three checks of a minute each, exactly what goes to the server, and what the service cannot do. Do read the limits: you are responsible for the result, not us.

What goes where

Three checks

  1. Go offline

    Open the site and let the page load. Turn off Wi‑Fi or switch on airplane mode, open a document and download the result. Everything works offline, so the server doesn’t need your file. For PDF, first open any PDF while online: its libraries load once.

  2. Watch the network

    Press F12, open the Network tab, clear the list, open a document and download the result. No request will carry your file. The only one is a tiny POST to /api/e: open it and look at the Payload, it has three fields: event type, format and language.

  3. Read the code

    All the code runs inside the page and is not minified. Open Sources, the jasyr.….js file in assets, and search for sendBeacon: there is one call to our server, the counter, and no fetch or XMLHttpRequest in the service’s own code (third-party libraries aside). To switch that off as well, enable Global Privacy Control in your browser and nothing is sent at all.

What we do receive

Never received or stored

  • document contents and redaction results
  • file names
  • replacement keys, your custom words and rules
  • cookies and accounts: there is no sign-up

Received and stored

  • daily counters: opens, downloads, formats (Word, PDF, Excel…), page language
  • unique visitors per day: from an irreversible fingerprint of IP and browser with a salt that changes daily; the IP itself is not stored

Caveats: our hosting provider keeps ordinary technical request logs, and fonts load from Google Fonts, so Google sees your IP.

What Jasyr cannot do

It is not an AI and can miss things

Rules reliably find structured data: IINs, phones, accounts, typical names. A rare name without context, a typo, an unusual notation or non-standard format can slip through. Always review the result.

Images and scans are not read

Stamps, signatures, photos and scans stay as they are. PDFs without a text layer are not supported. PDF pages are redrawn, so the original text and metadata do not reach the result, but images on them remain visible. Form fields, notes and stamp annotations are not carried over at all.

Not everything inside a file is processed

Word: body text, tables, text boxes, headers and footers, footnotes and comments. Deleted revisions are dropped; picture captions, bookmarks, file properties and the author are cleared. Excel: cells on every sheet, hidden ones included, notes, headers and footers, and sheet names. Check charts and text inside images by hand.

A person can be recognised without a name

A job title, a rare date, an unusual amount or the circumstances can point to a person even after the names are replaced. If the risk is high, hide amounts and dates and remove unique details.

Pseudonyms are reversible, blackout is not

The replacement key is a “Person-1 → real name” table, so it is a sensitive file in itself. Keep it apart from the document, never send them together and never send it to an AI. The black-bar mode is irreversible.

A tool, not a legal guarantee

Jasyr is not certified and does not replace a lawyer’s view under Kazakhstan’s personal data law. The decision whether a document may be sent is yours.

Before you send it to an AI

Open Jasyr