Data Leak Prevention for LLMs

Open source browser extension that detects sensitive data before you submit it to ChatGPT, Claude, or other public AI tools.

Catches emails, API keys, credentials, PII, and other patterns in real-time. All processing happens locally in your browser. No data collection, no tracking, no account required.

Install Extension

Compatible browsers:

Chrome

Edge

Brave

Opera

Vivaldi

The Problem

People paste sensitive information into ChatGPT and other public LLMs regularly. Company policies exist, but they're not enforced technically. Mistakes happen.

When something leaks, the person who pasted it is accountable — whether it was intentional or not.

Common data types that get accidentally pasted:

API keys, tokens, AWS credentials
Internal source code and system logs
Customer PII, PHI, and financial data
Client emails and confidential documents
Internal communications and strategy notes

LLM Guard adds a technical layer between you and the mistake.

Who Uses This

Software Engineers & DevOps

Prevent accidental exposure of code, credentials, architecture details, or system logs.

Security & IT Professionals

Lightweight DLP tool for teams that use public AI tools. Can be recommended as a baseline control.

Consultants & Contractors

Maintain client confidentiality and NDA compliance when using AI for work deliverables.

Compliance & Operations

Use AI tools for drafting and analysis while reducing risk of PII or PHI exposure.

How It Works

Extension monitors input fields on LLM sites

Runs locally in your browser. Content script watches for form submissions.

Text is scanned against detection rules

Regex patterns and keyword matching for emails, API keys, credentials, PII, and other sensitive data types.

Warning shown if sensitive data detected

Submission is blocked. You can review what was detected and choose to proceed or edit.

Event logs stored locally

Metadata only (timestamp, rule triggered, domain). The actual text content is never logged.

Fully configurable rules

Enable/disable detection patterns, add custom keywords or regex, adjust severity levels.

Technical Details

🔒

Local Processing Only

All scanning happens in-browser. No external API calls. Zero telemetry or analytics.

📋

10+ Built-in Rules

Detects emails, phone numbers, SSNs, credit cards, API keys, JWT tokens, AWS keys, IP addresses, and custom keywords.

⚙️

Configurable Detection

Add custom regex patterns, keywords, and exceptions. Adjust rule severity and enable/disable per rule.

📁

Metadata-Only Logging

Local logs store timestamp and rule triggered. Actual text content is never saved.

🌐

Multi-Platform Support

ChatGPT fully supported. Claude, Gemini, Perplexity, and Copilot in development.

FAQ

Does this send any data externally?

No. All processing happens locally in your browser. No analytics, no telemetry, no external API calls. Open source — you can verify this yourself.

What data does it log?

Only metadata: timestamp, rule triggered, and domain. The actual text content is never logged. Logs are stored in browser local storage, not sent anywhere.

Does it block submissions or just warn?

It blocks the submission and shows a warning. You can then choose to edit your input or proceed anyway. You stay in control.

Can I customize the detection rules?

Yes. You can enable/disable specific patterns, add custom regex rules, add custom keywords, and adjust severity levels. Full configuration UI included.

What detection patterns are included?

10+ patterns including: email addresses, phone numbers, SSNs, credit cards, API keys, JWT tokens, AWS credentials, private keys, IP addresses, and configurable keywords.

Which AI platforms are supported?

ChatGPT is fully supported. Adding Claude, Gemini, Perplexity, and Microsoft Copilot support soon.

Is there an enterprise version?

In development. Will include centralized management, team policies, reporting, and SSO. Individual version stays free. Join waitlist.

Can security teams deploy this?

Currently requires manual install per user. Enterprise version will support centralized deployment and policy enforcement.