Use AI safely without leaking your private data.
The LogosGuard browser extension runs locally in your browser and requires no account. When you paste content with sensitive data into AI tools like ChatGPT and Claude, it swaps the real values for placeholders (`[NAME]`, `[ACCOUNT]`, `[ADDRESS]`) before the prompt is submitted. The AI sees a working prompt and gives a useful answer. Your real values never leave your device.
You already use AI for everything: drafting emails, debugging code, reviewing contracts. The question is no longer whether to use AI, it's what you paste into it.
How redaction actually works.
When a prompt contains sensitive data, LogosGuard swaps the real values for placeholders before the prompt leaves the device. The AI sees something like `Help me draft a reply to [CUSTOMER]. Their account [ACCOUNT] is showing the wrong balance.` and produces a useful response against those placeholders. You get the AI's answer back and continue your work. The real customer name, account number, and other identifiers never leave your device.
This is the difference between LogosGuard and trying to remember what's sensitive. The AI tool is still useful. You don't have to slow down. Real data does not have to leave your device for the AI to be helpful.
What sensitive data LogosGuard detects.
Names
Full names and the contact names in your prompt.
Email addresses
Personal and work email addresses.
Phone numbers
Mobile, work, and other phone numbers.
Addresses
Street and mailing addresses.
URLs
Links and web addresses in your prompt.
Dates
Dates that can tie back to a specific person or event.
Account & customer IDs
Account numbers, customer IDs, and reference numbers.
Secrets
API keys, passwords, and access tokens.
What the extension catches before submit.
- Paste: clipboard content scanned the moment it lands in an AI tool's input field.
- Type: content scanned as you type, with detection running at typing speed.
- File upload: uploads scanned in supported formats (text, common document types) for the same patterns.
How it works in your flow.
- 1
Highlight
LogosGuard highlights the sensitive spans in your prompt before submission, so you can see exactly what was flagged.
- 2
Suggest
The extension proposes a redacted version of your prompt with placeholders in place of the real values.
- 3
Review
You decide. Accept the redaction, edit the prompt yourself, or send the original. You're always in the loop.
- 4
Submit
The AI sees the prompt you approved. Real values stay on your device.
Privacy posture.
- Local detection. Scanning happens in your browser. Your prompt content is not sent to LogosGuard servers as part of detection.
- No account required. No sign-up, no email, no login. Install from the Chrome Web Store and you're done.
- No server-side history. LogosGuard does not maintain prompt or scan history on its servers.
- Local model assets. The extension downloads its detection model from Hugging Face once and caches it locally. The download does not include your prompts or files.
- Local extension storage. Settings and state are stored in your browser. Removing the extension or clearing browser extension storage clears it.
- See the privacy policy for the full handling description.