Paste into AI. Keep private details private.
Find and redact sensitive details before your text reaches ChatGPT, Claude or another AI.
- Processed locally
- No text uploads
- No account needed
- 1 Paste text
- 2 Review matches
- 3 Copy safe version
Original text
Safe version
Paste something to scan it. It never leaves this tab.
Detection settings 7 rules on · placeholders
What it catches
Eight rule groups, all hard-coded in the page you're reading. Every pattern is readable source, so no server decides what counts as sensitive.
Email addresses
[email protected] is the fastest way to tie a paste back to a real person.
Phone numbers
US and international formats with separators, like +44 20 7946 0958.
API keys & tokens
20+ known shapes: OpenAI sk-…, GitHub ghp_…, AWS AKIA…, Stripe, Slack, Google and more.
Private key blocks
Anything between BEGIN and END PRIVATE KEY, a classic stowaway in pasted error logs.
JWTs
Session tokens hiding in curl commands and HTTP headers.
Internal hosts & IPs
db01.internal, 10.x, 192.168.x. Your network map is confidential too.
Password assignments
Lines like password = "…" in configs and scripts. The key name stays, the value goes.
Credit card numbers
Luhn-checked so order numbers don't false-alarm. Off by default, switch it on when you need it.
Every rule is a visible switch, not a black box. If you want the AI to see your email, leave that switch off. You're in charge.
Read this before you send
What PromptScrub cannot catch
Pattern matching finds shapes, not meaning. It will never know that “Project Falcon” is confidential. Skim your text one last time for:
- Real names of clients, colleagues or candidates
- Unreleased numbers such as revenue, growth, headcount or pricing
- Project code names your company treats as confidential
- Anything under NDA: contracts, negotiations, legal matters
- Other people's private details they didn't agree to share
Why your text stays local
PromptScrub runs the detection and redaction code in this browser tab. Your text is not sent to our server.
Scanning begins after you type or paste and finishes on this device.
The tool has no upload form or API route that receives the text you enter.
Open DevTools, watch the Network tab, then paste text. No request carries it away.
Practical guides
What not to paste into AI, and safer ways to get the same work done.
What a redaction tool cannot see in your text
Eight rule groups and 28 patterns can reliably find supported API key formats but cannot identify a confidential project name. The distinction is not a gap in the rules; it is the boundary of the whole method.
Your .env file hides its secrets from most redaction tools
DB_PASSWORD and PASSWORD look the same to a person and completely different to a regular expression. We shipped this bug ourselves, and here is the one character that caused it.
The fifteen-second check before you hit send
Automated rules handle the formats. This is the part only you can do, in the order that catches the most for the least effort.
Frequently asked questions
Why placeholders instead of just deleting?
Numbered placeholders like [EMAIL_1] keep the structure of your text so the AI still understands who did what. The same value always gets the same number, so references stay consistent. Deleting is safer when structure doesn't matter; you choose.
Will it catch everything?
No tool can. Rules catch well-defined shapes: emails, key formats, card numbers. They can't know that “Project Falcon” is confidential. That's what the five-point checklist above is for.