Detect
An on-device AI model reads your draft the moment you type it. No cloud call, no round-trip.
On-device privacy for AI prompts
Redact what AI shouldn't see.
PrAIvacy catches the names, emails, secrets and more in your ChatGPT, Claude and Gemini prompts, then redacts them in one click before you hit send.
On-device · no accounts · never sent to a server
Your message
Draft a follow-up to Jane CohenNAME at jane.cohen@acme.comEMAIL about the Q3 renewal. Her direct line is (415) 555-0142PHONE.
3 findings redacted
Send ↵
Works where you already prompt
ChatGPT
Claude
Gemini
Copilot
Perplexity
DeepSeek
Grok
Meta AI
HuggingChat
Qwen
Kimi
How it works
Three steps, all on your machine.
A privacy model runs inside your browser and reads your draft as you type. The scanning never touches a network.
Flag
Sensitive spans get underlined and ranked by severity, from a stray name to a leaked secret key.
Redact
One click swaps each finding for a safe placeholder before you send. Restore your originals whenever you need them.
On-device by design
Your prompts are scanned on your device.
Most "privacy" tools send your text to a server to scan it. PrAIvacy scans it right on your device instead.
Runs on your device
The detector is a private AI model that downloads once and then works fully offline.
Your text isn't sent to a server
The scan runs on your device. PrAIvacy never uploads your text, to us or anyone.
No account, no telemetry
No sign-up, no tracking of what you write. Two Chrome permissions and the AI sites you choose. That's it.
FAQ
Questions, answered.
Is PrAIvacy free?
Yes. PrAIvacy is free and runs entirely on your machine. No account, no subscription.
What data does it collect?
None. Detection runs on your device, and PrAIvacy has no server that receives your text. No accounts, no analytics, no tracking. See the privacy policy.
Does it need an internet connection?
Only once. The first time you turn on protection, PrAIvacy downloads its detection model. After that, every check runs on your device, fully offline.
Can I add my own words?
Yes, and it's free. Select any term in the prompt box — a client name, a project codename, an internal system — and flag it. From then on it's underlined everywhere, whatever the model thinks. Your list stays on your device.
Which AI sites does it work on?
ChatGPT, Claude, Gemini, Copilot, Perplexity, DeepSeek, Grok, Meta AI, HuggingChat, Qwen and Kimi. It only touches those sites, never every website.
Which browsers are supported?
Chrome and other Chromium browsers like Microsoft Edge.
Can I get my original text back?
Yes. Redactions are reversible. Restore your original wording anytime, right where you typed it.
Stop pasting secrets into AI.
PrAIvacy is free and runs entirely on your machine.
Free · coming soon to the Chrome Web Store