ClauseRadar — Privacy Policy

Effective Date: August 28, 2026
Version: 1.2
Extension Name: ClauseRadar
Developer: Halim Ashraf
Contact: Limpohal@users.noreply.github.com

1. Overview

ClauseRadar ("we", "us", "our") is a Chrome browser extension designed for regulatory affairs professionals to classify the significance of regulatory clauses and compare document versions. This Privacy Policy explains what information we collect, how we use it, and your rights regarding your data.

We are committed to protecting your privacy. We collect only the minimum data necessary to provide the service, and we do not sell your data to third parties.

2. Information We Collect

2.1 Data You Provide Voluntarily

Clause text you paste or select — Text you paste into the extension or highlight on web pages for classification/comparison
Framework selections — Your chosen regulatory frameworks (EU MDR, FDA, ISO, etc.)

2.2 Automatically Generated Data

Anonymous Device ID — A randomly generated UUID (e.g., 550e8400-e29b-41d4-a716-446655440000) created on first install and stored locally in chrome.storage.local. This is used only for rate limiting (20 requests/day per device, plus a global daily service-capacity limit shared across all users to control infrastructure costs).
Usage Timestamps — When you make a classification or comparison request (for rate limit window calculation).
Global Service-Capacity Counter — A daily, date-keyed counter of total AI model calls across all users (a number only — no device ID, no clause text). Enforces the service-wide capacity limit that protects against abuse and keeps the service free.

2.3 Data We Do NOT Collect

❌ No personally identifiable information (name, email, IP address, location)
❌ No browsing history or web activity
❌ No authentication credentials or account data
❌ No clipboard contents beyond what you explicitly paste into the extension
❌ No analytics, tracking, or advertising identifiers

3. How We Use Your Data

Data Type	Purpose	Legal Basis
Clause text	Send to backend for AI classification/comparison via OpenRouter	Legitimate interest (providing requested service)
Framework selection	Tailor AI prompts to selected regulatory context	Legitimate interest
Device ID	Enforce daily rate limit (20 requests/day)	Legitimate interest (abuse prevention)
Global capacity counter (anonymous number)	Enforce the daily service-wide AI-call budget	Legitimate interest (abuse prevention, cost control)
Timestamps	Calculate 24-hour rate limit windows	Legitimate interest

We do not use your data for: advertising, profiling, marketing, selling to third parties, or any purpose beyond providing the classification/comparison service.

4. Data Sharing & Third Parties

4.1 Primary Backend (Cloudflare Workers)

Provider: Cloudflare, Inc.
Data sent: Clause text, framework selection, anonymous device ID
Purpose: Process classification/comparison requests, enforce rate limits
Retention: Request counters expire after 2 days. Clause text is not stored server-side beyond the request lifecycle.

4.2 Fallback Backend (Vercel)

Provider: Vercel Inc.
Data sent: Clause text, framework selection, anonymous device ID
Purpose: Process classification/comparison requests as a fallback when the primary backend is unreachable (e.g., corporate network restrictions). The extension automatically falls back to Vercel if the Cloudflare Worker fails.
Retention: Rate limit counters are stored in-memory only and reset when the serverless function is recycled. Clause text is not stored beyond the request lifecycle.

4.3 AI Model Provider (OpenRouter)

Provider: OpenRouter (openrouter.ai)
Data sent: Clause text + framework-specific prompt (no device ID, no user identity)
Purpose: Generate classification/comparison results
Policy: See OpenRouter Privacy Policy
Note: We use the openai/gpt-5.6-luna model via OpenRouter. OpenRouter acts as a gateway; we do not control their sub-processors.

4.4 No Other Sharing

We do not share data with any other third parties, advertisers, data brokers, or analytics providers.

5. Data Storage & Retention

Storage Location	Data	Retention
Local (`chrome.storage.local`)	Device ID, draft text, history (last 50 items), framework preferences	Until you uninstall the extension or clear browser data
Primary Backend (Cloudflare KV)	Daily request counter per device ID + global daily AI-call counter (anonymous)	2 days (auto-expires)
Fallback Backend (Vercel — memory)	Daily request counter per device ID + global daily AI-call counter (anonymous)	In-memory, resets on function recycle (no persistent storage)
Backend (Memory — both providers)	Clause text during request processing	Milliseconds (request lifecycle only)
OpenRouter	Per their policy	Per OpenRouter Privacy Policy

You can clear all local data anytime: Chrome → Extensions → ClauseRadar → Details → "Clear storage" or use the "Clear history" button in the extension's History tab.

6. Your Rights

Under applicable privacy laws (GDPR, CCPA, etc.), you have the right to:

Access — Know what data we hold (this policy + local storage inspection)
Delete — Uninstall the extension or use "Clear history" to remove all local data
Portability — Export your history as Markdown via the "Copy as Markdown" / "Download .md" buttons
Object/Restrict — Disable the extension to stop all processing
Human oversight of AI output — The AI suggests classifications (Critical/Minor/Administrative), but you retain full control to accept, override, or export results

To exercise these rights, simply use the extension's built-in controls or uninstall it. No account or request process is needed.

7. Children's Privacy

ClauseRadar is not directed to children under 13 (or 16 in the EU). We do not knowingly collect data from children. If you believe a child has provided data, contact us and we will delete it.

8. Security

All network communication uses HTTPS (TLS 1.2+)
No credentials or API keys stored in the extension (OpenRouter key is server-side only — set as environment variables on Cloudflare Workers and Vercel)
API keys are stored as platform secrets (wrangler secret / Vercel environment variables), never committed to source code
Device ID is a random UUID, not derived from hardware or personal info
Content Security Policy restricts extension scripts to self and known backend endpoints only

9. International Transfers

Your clause text is sent to:

Cloudflare Workers (global edge network — data may process in any Cloudflare data center)
Vercel (global edge network via AWS/Lambda — US and EU regions)
OpenRouter (US-based, may use global model providers)

By using the extension, you consent to this transfer. All providers maintain appropriate safeguards.

10. Changes to This Policy

We may update this policy occasionally. Material changes will be noted in the extension's changelog or via a notification. Continued use after changes constitutes acceptance.

11. Chrome Web Store Compliance

This extension complies with the Chrome Web Store Developer Program Policies:

Single Purpose: Clear, narrow functionality (regulatory clause classification/comparison)
Minimal Permissions: Only contextMenus, sidePanel, storage, activeTab, scripting
Narrow Host Permissions: Only https://clauseradar.ahali-clauseradar.workers.dev/* and https://vercel-fallback-neon.vercel.app/*
No Deceptive Practices: No hidden behavior, no data selling, no misleading UI
Privacy Policy: This document, accessible via store listing

12. Contact

For questions, concerns, or data requests:

Email: Limpohal@users.noreply.github.com
GitHub: https://github.com/Limpohal/ClauseRadar

13. Summary for Users

What Happens to Your Clause Text?
1. You paste/select text in the extension
2. Text + framework choice sent to our backend (HTTPS)
3. Primary Cloudflare Worker handles the request (auto-failover to Vercel fallback if unreachable)
4. Backend checks daily limit (anonymous device ID)
5. Backend forwards text + prompt to OpenRouter (using `openai/gpt-5.6-luna`)
6. AI returns classification/comparison
7. Result shown to you; text discarded from server
8. Only a counter (device ID + date) stays in KV (Cloudflare) or memory (Vercel) for up to 2 days

You own your data. We process it only to give you the result you asked for.