PII and secret scrubber

Hide personal details and secrets in text before you share it with an AI assistant, a forum or a support desk.

Hide

This free PII redaction tool works in your browser. Everything runs in your browser. Nothing you type is sent anywhere or stored. How the files you choose are handled is set out in the privacy policy.

Why scrub text before using AI

Whatever you paste into an AI chat, a public forum or a ticketing system may be stored, reviewed by staff or used for training. Logs and documents often hold more personal data than you notice: customer emails, phone numbers, IP addresses and live API keys. Replacing them with labels keeps the meaning, so the assistant can still help, without exposing anyone.

Consistent labels

Each different value gets its own numbered label, and the same value always gets the same label. If one email appears five times it becomes [EMAIL_1] five times, so the AI can still follow who is who. The list of replacements stays on this page, so you can map the answer back.

Limits

  • Detection uses patterns, so unusual formats can be missed. Read the result before sharing.
  • Names cannot be found automatically; add them in the box.
  • Card numbers are only hidden if they pass the Luhn check, which avoids hiding order numbers.

Worked example: Redacting customer emails, IPs, and JWTs from a support log before AI debugging

Suppose you want to paste a failing production error log into an AI coding assistant or a shared Jira ticket without leaking customer PII or live bearer tokens:

2026-11-14T09:12:04Z ERROR Auth failed for user priya.sharma@example.org (IP 203.0.113.45)
Phone on file: +91 98765 43210 | Retry by priya.sharma@example.org
Header: Authorization: Bearer eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJzdWIiOiI5OSJ9.sig123

When you run the PII and Secret Scrubber with consistent numbered tokens enabled, both occurrences of priya.sharma@example.org map to the exact same [EMAIL_1] placeholder so you can still trace that the same user retried:

2026-11-14T09:12:04Z ERROR Auth failed for user [EMAIL_1] (IP [IP_1])
Phone on file: [PHONE_1] | Retry by [EMAIL_1]
Header: Authorization: Bearer [JWT_1]

Limits and what this tool cannot do

  • Pattern-based regex detection (cannot catch unstructured human names or street addresses): The scrubber reliably matches structured patterns—emails, phone numbers, IPv4/IPv6 addresses, credit card numbers (Luhn check), SSN/PAN/Aadhaar patterns, AWS keys (AKIA...), and JWTs (eyJ...). It does not run a natural-language NER model, so plain prose like "Spoke with Rahul at 42 MG Road" must be reviewed manually.
  • Always inspect the output before sharing externally: Treat automated scrubbing as a fast first pass, not a substitute for reading the final snippet before pasting it into an external service.

Questions

Is my text sent anywhere?

No. The scrubbing happens in your browser. That is the point.

Which API keys are detected?

OpenAI and Anthropic keys, AWS access keys, GitHub, GitLab and Slack tokens, Google API keys, Stripe keys, JWTs, private key blocks, and values after words like password= or token=.

Is this enough for GDPR or DPDP compliance?

It helps, but it is not a guarantee. Check the result and follow your organisation’s data policy.

Why does the scrubber use numbered labels like [EMAIL_1] and [EMAIL_2] instead of [REDACTED]?

When debugging a multi-line log or email thread involving two or three different users or IP addresses, numbered placeholders let you see whether two events came from the same user ([EMAIL_1]) or a different user ([EMAIL_2]) without revealing who they are.

Which secret and API key formats are detected automatically?

It detects JSON Web Tokens (eyJ...), AWS Access Key IDs (AKIA... / ASIA...), GitHub personal access tokens (ghp_... / github_pat_...), Slack tokens (xoxb-...), Stripe keys (sk_live_...), and PEM private key blocks.

Is my unredacted text sent anywhere during scrubbing?

No. Every regular-expression pass runs inside your browser's local memory and nothing is logged or transmitted.

Last reviewed .

Related converters and guides