Prompt Sanitizer
Clean it before the AI sees it
People paste contracts, logs, emails, and support tickets into AI chatbots every day — often with names, addresses, card numbers, and API keys still in them. The moment that text reaches a cloud model it has left your control and may be logged, retained, or reviewed. This tool strips the sensitive parts first, on your device, so what you paste is already clean.
What it finds
It targets structured personal data, which is both the most dangerous to leak and the most reliably catchable: email addresses, phone numbers, credit-card numbers (validated with the Luhn checksum), national-ID numbers, IP addresses, URLs, physical addresses, and long secret tokens like API keys. Each match becomes a clear placeholder such as [EMAIL], so the text still reads sensibly and the model still follows it.
Labels or solid blocks
Labelled placeholders let the AI understand the structure of your text without seeing the real values — usually what you want for a helpful answer. Solid blocks hide even the type of data that was there, which suits text you are sharing with people or when the category itself is sensitive. You choose per use.
A growing need — with a caveat
The more you and your team rely on AI, the more sensitive text flows through it, and the bigger the leak risk — so sanitising first is a simple habit that only becomes more important, and it is already a requirement in many workplaces. One caveat: pattern matching cannot know context, so plain names or one-off identifiers can slip through. Always give the cleaned text a quick read before you send it.
Frequently asked questions
Does my text get uploaded?
No — the whole point is the opposite: the text is scanned and cleaned on your device, before it ever goes to a cloud AI. Nothing is sent to a server here.
What does it detect?
Emails, phone numbers, credit-card numbers (validated with the Luhn checksum), national IDs like US SSNs, IP addresses, URLs, physical addresses, and long secret-looking tokens such as API keys.
How does it hide the info?
It replaces each match with a clear label like [EMAIL] or [PHONE], or with a solid block — your choice — while leaving the rest of the text intact.
Should I use labels or solid blocks?
Use labelled placeholders when you still want the AI to understand the structure of your text; use solid blocks when you want to hide even the type of data that was there, or when sharing the text with people.
Is it perfect?
It catches well-structured data reliably, but no automatic tool catches every name or edge case — plain names and one-off identifiers have no fixed pattern, so always glance over the result before you paste it.
Can I rely on this for compliance?
It is a practical safeguard that reduces what you expose, not a certified compliance product — combine it with your own review and your organisation’s policies for regulated data.