on-device · browser-local · nothing leaves your machine

Don't ship your
secrets to the prompt box.

SafeHai watches what you paste into ChatGPT, Claude, Gemini and any Chrome form — and steps in before your Aadhaar, PAN, AWS keys or bank details go out with it.

safehai-extension.zip · 1.1 MB · Manifest V3 · free, no account needed

chat.ai — compose

Hi, so I want you to review my details —

0 leaks caught
in this demo
10data categories
detected
100%local scanning —
nothing uploaded
0accounts or sign-ups
required
<1msto flag text before
you hit send

$ cat detected_categories.log

Built to recognize what actually leaks

Every category below is matched by on-device pattern rules — the moment it shows up in a text box, SafeHai flags it before the page ever sees it submitted.

ID

Aadhaar number

12-digit national ID patterns, including spaced and grouped formats.

PAN

PAN card

Indian tax ID format — letters, digits and checksum position.

AWS

AWS access key

AKIA-prefixed access keys and paired secret strings.

sk-

OpenAI / API keys

sk- style tokens for OpenAI, Anthropic and other LLM providers.

gh_

GitHub token

ghp_ and fine-grained personal access tokens.

-----

Private key

PEM-format RSA / SSH private key blocks.

****

Payment card

Card numbers with valid Luhn checksums, any major issuer.

@upi

UPI ID

handle@bank style payment identifiers.

•••

Passwords

Credential-shaped strings typed next to login-style context.

!!

Phishing sites

Look-alike and known unsafe domains, flagged as you land on them.

$ ./how-it-works.sh

Three steps, all inside your browser

01

It watches quietly

Once installed, SafeHai runs as a content script on every tab — Gmail, ChatGPT, Claude, random web forms — sitting in the background, doing nothing until it sees something risky.

02

It flags before you send

The moment sensitive text lands in an input box, a warning drops in with exactly what was found — "Indian phone number," "AWS access key" — no vague alerts, no guessing.

03

You choose what happens

Redact & continue masks the sensitive part and sends the rest. Send anyway lets you override when you know it's fine. Either way, the decision stays yours.

$ ls ./screenshots

See it live

Real captures of SafeHai running in Chrome — the popup dashboard, the in-page warning, and a prompt-box catch.

SafeHai's in-page panel overlaying a website, showing zero leaks blocked today across all tracked categories.
In-page protection panel. A live count of what's been blocked, broken down by category, without leaving the page you're on.
An AI chat input box with a warning that reads one sensitive item detected, Indian phone number, with options to redact all and continue or send anyway.
Caught mid-prompt. One tap redacts the sensitive span and lets the rest of your message through.

$ unzip & load --unpacked

Install it in under a minute

No Web Store review queue yet — you load it straight from the folder. It's the same handful of clicks every developer extension asks for.

Download safehai-extension.zip
  1. 1

    Download & unzip

    Grab safehai-extension.zip above and extract it anywhere on your computer — remember the folder.

  2. 2

    Open Chrome extensions

    Go to chrome://extensions in the address bar.

  3. 3

    Turn on Developer mode

    Flip the Developer mode toggle in the top-right corner of the page.

  4. 4

    Load unpacked

    Click Load unpacked and select the unzipped SafeHai folder.

  5. 5

    Pin it

    Click the puzzle-piece icon in the toolbar and pin SafeHai so it's always one click away.

on the roadmap

Next: a local LLM, running on your device

Pattern rules catch known formats today. We're adding a small on-device model so SafeHai can reason about context too — no data ever has to leave your browser to get a second opinion.

$ git log --authors

Built by

Team Leader
S

Shashikant

J

Jayansh Khatri

A

Ayush Sharma

M

Mihir

Stop the leak before it leaves your keyboard.

Download SafeHai