Does my prompt get sent anywhere?

No. DataBlur AI runs 100% locally in your browser — there are no network calls, no analytics and no telemetry.

Which AI sites does it work on?

ChatGPT, Claude, Gemini, Copilot, Perplexity, Mistral, Grok, Poe and DeepSeek out of the box — and you can add any other site from the popup.

Will it block my messages?

Only if you want it to. In Warn mode it pauses on send so you can decide; Auto-redact cleans the prompt for you; Highlight-only never blocks.

Can it miss something?

Detection is pattern-based and tunable from strict to lenient. It is a strong safety net, not a guarantee — you always see and control what gets redacted.

How much does it cost?

The extension is free. It installs and protects your prompts at no cost.

100% Local | No Cloud | No AI

Redact secrets before they reach the AI.

DataBlur AI watches the prompt box on ChatGPT, Claude, Gemini and more — and steps in before a sensitive prompt is ever sent.

chatgpt.com

What can I help with?

Works on your AI tools

ChatGPT

Claude

Gemini

Copilot

Perplexity Mistral

Mistral

Grok

Poe

DeepSeek

ChatGPT

Claude

Gemini

Copilot

Perplexity Mistral

Mistral

Grok

Poe

DeepSeek

Three modes

You choose how careful it is.

Pick one default and change it per site whenever you like.

My name is Jonathan Smith, my SSN is 123-45-6789, email leak@corp.com, and the key is sk-ant-api03-AbcDef123456789.

4 flagged

Highlight only

Never block or alter a send — just flag what looks sensitive as you type.

Sensitive data detected

ssn123-45-6789

emailleak@corp.com

nameJonathan Smith

Send anywayRedact & Send

Warn before send

Pause on send and let you edit, redact & send, or send anyway. The default.

My name is [NAME_1], my SSN is [SSN_1], email [EMAIL_1], and the key is [SECRET_1].

I only received placeholders — your real name, SSN, email and key never left your browser.

Ask anything

Auto-redact on send

Redact flagged values automatically, with a one-click Undo toast right after.

What it catches

Six kinds of secret, flagged as you type.

Each category toggles independently, with a strict / balanced / lenient sensitivity dial.

Contacts

Emails and phone numbers.

Financial

Card numbers and IBANs.

Secrets

API keys, tokens and passwords.

Identity

Names, national IDs and passports.

Crypto

Wallet addresses and seed phrases.

Money

Amounts and account numbers.

Private by design

Nothing ever leaves your browser.

Every scan happens locally. There is no server to send your prompts to.

Plans

Free for you, five dollars a seat for your team.

Both plans land with the first release. Nothing is billed today.

What you get	Personal	Teams
Highlight, warn and auto-redact modes
Six detection categories with a sensitivity dial
100% local detection, nothing leaves the browser
Every supported AI site
Custom regex rules	On device	Pushed org-wide
Admin dashboard
Central rule and policy management
Locked mode per site, enforced for everyone
SSO (SAML) and SCIM provisioning
Managed rollout via Chrome / Edge policy
Site allow and deny list
Redaction audit log — counts only, never content
Priority support with an SLA
Price	Freefor personal use	$5per seat / month

Pricing is not final and no card is taken yet. Tell us what your team needs and it can still make the first release.

FAQ

Questions, answered.

05 / Get DataBlur AI

Keep your prompts to yourself.

100% Local | No Cloud | Free