Does masking change my original file?
No. The masker makes a new copy in your browser tab. The original text or file is not modified, and nothing is saved unless you download the copy.
Make a masked copy of any text before it goes into a ticket, a chat, a log excerpt, test data, or an AI prompt. Choose per category whether to keep the last four digits, replace everything, write a label such as [EMAIL], or use consistent tokens so the masked text still makes sense.
Loading the PII tools.
You do not have to take our word for it. Browser developer tools list every request a page makes, so you can watch 2pii work and confirm that your text is never sent anywhere. The check takes about a minute.
F12, or Cmd+Option+I on a Mac, and select the Network tab.Text you paste or open is held in this tab's memory, checked by code that is already on the page, and handed back to you as a copy or a download. It is never uploaded, never stored in the browser, and gone when you close the tab. The only other request the site can make is the feedback form, and only when you submit it.
The production site also sends a Content-Security-Policy header that tells your browser to refuse connections to anywhere except this site and the anonymous analytics endpoint, so even a bug could not send your data elsewhere.
***-**-1234, **** **** **** 4242), the first letter and domain of an email (j*******@example.com), the year of a birth date (**/**/1985), the state and first three ZIP digits of an address, the first two parts of an IP address, and the initials of a name.***-**-****).[SSN] or [EMAIL].[EMAIL-1]. The same value always gets the same token, even when it is written differently (a phone number with or without dashes), so a masked conversation or log still shows who is who.The mask character can be an asterisk, a capital X, or a bullet.
Tokens are the best choice when the masked text will be read for meaning. A support thread, an incident timeline, or a prompt to an AI assistant often depends on knowing that the same customer or the same card appears several times. Asterisks hide that; tokens keep it without keeping the value.
Partial masking is best for audit evidence and tickets where an analyst may need to match a record by its last four digits. Labels are best for documentation and examples, where the category matters and the identity does not.
Not by itself. Masking hides the values the scanner finds. De-identification is a stronger claim: that nobody can reasonably re-identify the people in the data, which depends on everything left in the text, including details the scanner cannot see. HIPAA, for example, recognizes only Safe Harbor (removing 18 kinds of identifiers) and Expert Determination as de-identification methods.
Use the masker to reduce exposure in everyday sharing. For a formal de-identification, review the whole data set against the method your law or policy requires, and see the HIPAA sources in the laws directory.
Whichever route you take, read the result before sending it. Masking only covers what the detectors found.
Your original text or file is never modified. The masker builds a new string in the tab's memory, and nothing is written to your disk unless you press Download. Values the scanner did not detect, such as an unlabelled name in a sentence, stay as they are, which is why the count of masked values is shown above the output.
Everything is on the 2pii home page. Part of the 2suite.io privacy-first ecosystem.
No. The masker makes a new copy in your browser tab. The original text or file is not modified, and nothing is saved unless you download the copy.
It replaces each value with a numbered placeholder such as [EMAIL-1], and the same value always gets the same token, so the masked text still shows which records match.
Yes. Each category found in the text has its own on/off switch and style, so you can, for example, keep phone numbers and mask everything else.
No. 2pii works on text. For PDFs and images, use 2redact, which permanently redacts the document in your browser.
Not automatically. It hides what the detectors found, but unlabelled names, free-text details, and context can still identify people. Read it first.