How to use PII scrubber
- Paste the text: a log, an email thread, a support ticket, a document.
- Choose what to find, and list any names or companies to hide as well.
- Pick how to replace them, then copy the scrubbed text.
Scrub before you share
Pasting a log file into an AI chat to debug it, forwarding a customer email to a contractor, posting an error to a public forum, or sharing a document with a vendor: each can leak personal details and secrets that weren't meant to go anywhere. Once text is in someone else's system, you can't take it back. This tool finds personal data and credentials in text and replaces them before you share it, so the useful part of the text survives and the sensitive part doesn't.
What it catches
Email addresses and phone numbers in international and national formats; payment card numbers, but only digit runs that pass the Luhn check cards use, so order numbers aren't mistaken for cards; IBANs, verified with their built-in check digits; US Social Security numbers and UK National Insurance numbers, skipping ranges that are never issued; IPv4 and IPv6 addresses; and secrets: AWS access keys, GitHub, Stripe, Slack, OpenAI, Anthropic and Google API keys, JSON Web Tokens, private key blocks, and passwords embedded in database or service links.
Names, companies and street addresses can't be recognised reliably by patterns, because they look like any other words. Add the ones that matter to the list of terms, and each is hidden wherever it appears, whatever its capitalisation.
Keep the meaning
Choose how matches are replaced. Numbered placeholders, such as EMAIL_1 and EMAIL_2, give the same value the same number every time, so a conversation or log still shows who wrote to whom and which account did what; that's usually best when asking an AI to analyse the text. Labels like [EMAIL] are simpler, and blocks of █ look like a classic redaction. A summary shows how many of each kind were hidden.
Private by design
A scrubbing tool that uploads your text would defeat its purpose. Here detection and replacement run in your browser as you type, and nothing is sent or stored. Still read the result before sharing: patterns catch the common formats, not every way a person can write their details.
Questions
- What does it find?
- Email addresses, phone numbers, payment card numbers (checked with the Luhn algorithm), IBANs (checked with their mod-97 digits), US Social Security and UK National Insurance numbers, IP addresses, API keys and tokens from AWS, GitHub, Stripe, Slack, OpenAI, Anthropic and Google, private keys, JWTs, and passwords inside links.
- Can it find names?
- Not by itself: names look like any other words, so pattern matching can't spot them reliably. List the names, companies and addresses that matter, and they're hidden wherever they appear, in any capitalisation.
- Is the text sent anywhere?
- No. That's the point: you scrub text before it goes to an AI tool or a third party, so the scrubbing itself happens in your browser and nothing is uploaded or stored.