Free for Chrome

Use AI.
Keep your secrets.The Chrome extension that hides your personal data from ChatGPT, Claude and Gemini.

It anonymizes your prompts automatically: names, emails, phone numbers, IDs and card numbers become placeholders before you hit Send. You still get a useful answer, and the real details never leave your browser.

Add to Chrome, it's free Watch the 45-second tour
  • Runs only in your browser
  • No account, prompts never collected
  • Free to start
your-ai-assistant.com SafePrompt on
AI Assistant

Works on

  • ChatGPT
  • Claude
  • Gemini
  • Grok
  • Copilot
  • DeepSeek

Once you hit Send, it's out of your hands.

It can be kept. Whatever you paste into an AI chat can be stored, reviewed by people, or used to train future models.

It's not just lawyers. "Email this to…", "my phone is…", "here's my ID…" Everyday prompts leak data just as easily.

At work it's someone else's data. One pasted contract, payroll line or patient note exposes a client or colleague, not only you.

3.8

pastes of sensitive company data per day, on average, by employees who paste into AI tools

~40%

of files uploaded to AI tools contain personal or payment card data

72%

of AI use at work happens on personal accounts, outside company controls

Source: LayerX, Enterprise AI and SaaS Data Security Report 2025.

See SafePrompt work in 45 seconds

A warning before you send, placeholders on the way out, your real details back on your screen.

Watch on YouTube

How SafePrompt keeps personal data out of AI chats

SafePrompt sits on the chat page you already use. You keep typing the way you do now.

  1. 1

    It warns you before you send

    Press Enter or Send as usual. If SafePrompt spots sensitive data, it lists what it found and lets you choose: Secure & send, Send anyway, or go back and edit.

    Sensitive data detected

    This message contains sensitive data. Secure & send replaces it with placeholders before sending.

    NameEmailPhone numberID
    Send anywaySecure & send
    Write a follow-up email to Jane Cohen (jane.cohen@acme.com, 054-381-2297) about invoice #4471 for client ID 3049-7731.
  2. 2

    Your data becomes placeholders

    One click and each detail is swapped for a token like [[SP_EMAIL_1]]. The AI gets the shape of your request, never the real values. Clicking twice won't double-mask.

    Secured!
    Write a follow-up email to [[SP_NAME_1]] ([[SP_EMAIL_1]], [[SP_PHONE_1]]) about invoice #4471 for client ID [[SP_ID_1]].
  3. 3

    Mask anything, just highlight it

    A nickname, a case name, a project codename. Select it and click Mask as PII. Every match in that prompt is replaced, including things auto-detect wouldn't know about.

    Mask as PII Summarize the Rosenberg case notes and draft next steps for the Rosenberg family.
  4. 4

    Reveal the real details, on your screen

    When the answer arrives, click Reveal Original Data. Placeholders turn back into the real names and numbers right on the page, so you can read and copy a finished draft.

    Subject: Invoice #4471, quick follow-up

    Hi Jane Cohen,

    Just following up on invoice #4471 for client 3049-7731. If it's easier, I can call you at 054-381-2297 this week to wrap it up.

    Reveal Original Data
Add to Chrome, it's free Takes about 10 seconds. Open ChatGPT and it's already working.

What personal data SafePrompt masks

Detection runs on every prompt, on every supported site, and works inside JSON, SQL and URLs without breaking them.

  • Names
  • Email addresses
  • Phone numbers
  • ID numbers, SSNs and Israeli IDs
  • Credit card numbers
  • IBANs and bank details
  • Home addresses
  • IP addresses
  • Anything you highlight

Private by design: runs 100% in your browser

SafePrompt protects your data from the AI provider, and it never sends it to us either.

  • Local detection. Finding and replacing data happens inside the extension. Your prompt text isn't sent to SafePrompt.
  • A vault that expires. The token map lives in Chrome's session storage for that tab. It expires after about 30 minutes, and clears when you close the browser or click Clear Session History.
  • Reveal stays on your screen. Swapping tokens back into the answer happens on the page you're looking at. It never leaves your machine.
  • Only two optional connections. A license check with Lemon Squeezy if you buy Pro, and a Google Form if you choose to report an issue.

Read the full privacy policy

Made for anyone who pastes real life into AI

Everyday use

Emails, forms, letters and messages that mention you, your family or your friends.

Legal and HR

Contracts, case notes, payroll lines and hiring plans, without the names attached.

Finance and accounting

Client IDs, IBANs, card numbers and invoices, kept out of the model.

Support and sales

Customer emails and phone numbers stay yours while the AI drafts the reply.

Developers

Emails and IPs inside logs, JSON and SQL get masked without breaking the syntax.

Healthcare and care work

Summarize notes without handing over who they're about.

SafePrompt pricing: start free, upgrade when it's part of your day

Warnings and detection are always free. Secure Send and Reveal come with 10 free uses, then Pro is $6.99 a month, or $56.95 a year.

Free

$0

For trying it on real prompts

  • Warnings before you send, always on
  • Detection on every supported AI site
  • Mask as PII for anything you highlight
  • 10 Secure Sends, Reveal included
Add to Chrome

Pro

$6.99 per month

or $56.95 a year (save 32%)

For people who use AI every day

  • Everything in Free
  • Unlimited Secure Sends
  • Unlimited Reveal Original Data
  • Activate with a license key in the extension
Upgrade to Pro

Rolling SafePrompt out to a team? Email us and we'll set you up.

Guides to using AI without giving away personal data

Straight answers about what happens to what you type, and how to keep names, numbers and client details private.

All guides

ChatGPT privacy questions, answered

Is it safe to share personal information with ChatGPT?

It's safest not to. On consumer plans, OpenAI says it may use your conversations to train its models unless you turn off "Improve the model for everyone" in Settings, Data controls. Even with training off, your chats are still kept in your account. SafePrompt replaces names, emails, phone numbers, IDs and card numbers with placeholders before you send, so the real values never reach ChatGPT, Claude or Gemini.

What should you never share with ChatGPT or other AI chatbots?

Anything that identifies a real person or account: full names with contact details, ID and passport numbers, credit card and bank details, home addresses, and other people's data from work, like client records, payroll lines or patient notes. SafePrompt warns you when a prompt contains these and masks them in one click.

How do I anonymize a prompt before sending it to an AI?

Install SafePrompt and type as usual. When you press Send, it lists the personal data it found and offers Secure & send, which swaps each detail for a placeholder like [[SP_EMAIL_1]]. You can also highlight any word and click Mask as PII. When the answer comes back, Reveal Original Data puts the real values back on your screen.

Can SafePrompt see my prompts?

No. Detection and masking run inside the extension in your browser. Your prompts and the AI's answers aren't sent to SafePrompt, and nothing is stored on a server.

Isn't it enough to just be careful?

Most leaks aren't decisions, they're habits: pasting a whole email thread, a contract or a log file without reading every line. SafePrompt checks at the moment that matters, when you press Send, and only interrupts when it finds something.

Will the AI still understand what I'm asking?

Yes. Placeholders keep the structure of your request, and SafePrompt adds a short line asking the model to keep the tokens unchanged. When the answer comes back, Reveal Original Data puts the real values in place.

Which AI sites does it work on?

ChatGPT, Claude, Gemini, Grok, Microsoft Copilot and DeepSeek, in Chrome.

Can I send something without masking it?

Yes. Choose Send anyway in the warning. It applies to that one prompt, and you'll be warned again next time. You can also pause protection from the extension popup.

What happens when I refresh or close the tab?

The token map is kept per tab for about 30 minutes so Reveal still works after a refresh. Closing the browser, or clicking Clear Session History, wipes it.

What's included for free?

Warnings, detection and Mask as PII are always free. You also get 10 Secure Sends, with Reveal included. After that, Pro unlocks unlimited use for $6.99 a month, or $56.95 a year.

Something wasn't detected, or was flagged by mistake. What do I do?

Highlight anything that was missed and click Mask as PII. To help us improve detection, use Report an issue in the extension or email safeprompt1@gmail.com.

Use AI. Keep your secrets.

Install SafePrompt and your next prompt is already protected.

Free on the Chrome Web Store

Works with ChatGPT, Claude, Gemini, Grok, Copilot and DeepSeek