Inspiration
Privacy policies are notoriously long, complex, and filled with legal jargon that most users never read. Studies show that 97% of people accept privacy policies without reading them, unknowingly agreeing to extensive data collection and tracking. We wanted to democratize privacy awareness by making it instantly accessible and understandable for everyone, empowering users to make informed decisions about their digital footprint without needing a law degree.
What it does
Safemode is a Chrome extension that acts as your personal privacy guardian. It automatically analyzes website privacy policies and cookie practices using Chrome's built-in AI to provide:
- Instant Risk Assessment: A 0-100 privacy risk score with SAFE, CAUTION, or UNSAFE verdicts
- Plain-English Summaries: AI-powered breakdowns of complex privacy policies
- Cookie Analysis: Detailed categorization of tracking cookies (essential, analytics, advertising)
- Data Rights Information: Clear explanations of your data deletion rights and how to exercise them
- Actionable Recommendations: Specific steps you can take to protect your privacy
- Auto-Scan Mode: Optional automatic analysis as you browse, with warnings for high-risk sites
All processing happens locally in your browser—no external servers, no data collection, complete privacy.
How we built it
We built Safemode as a Chrome Manifest V3 extension using:
- Chrome Built-in AI APIs: Leveraging Chrome's experimental summarizer and language model APIs for on-device analysis
- Vanilla JavaScript: Content scripts for web scraping, service workers for orchestration, and popup UI for user interaction
- Custom Privacy Scrapers: Pattern-matching algorithms to detect and extract privacy policy links and content
- Cookie Analysis Engine: Deep inspection of all cookies with intelligent categorization
- Risk Scoring Algorithm: Multi-factor assessment considering data collection, third-party sharing, deletion rights, and tracking intensity
- Caching System: 24-hour result caching to minimize API calls and improve performance
The architecture uses content scripts to scrape page data, a service worker to orchestrate AI analysis with structured prompts, and a clean popup interface to present results.
Challenges we ran into
- Experimental AI APIs: Working with Chrome's cutting-edge built-in AI required extensive experimentation and fallback handling for unsupported environments
- Privacy Policy Variability: Websites use wildly different formats, naming conventions, and structures for privacy policies, requiring robust pattern matching
- Accurate Risk Scoring: Developing a fair, consistent algorithm that balances multiple privacy factors without being too sensitive or too lenient
- Performance Optimization: Parsing large privacy documents and running AI analysis without blocking the browser or consuming excessive resources
- Testing at Scale: Creating comprehensive test cases that cover the wide spectrum of privacy practices across the web
Accomplishments that we're proud of
- 100% Local Processing: Achieved complete privacy-preserving analysis using only Chrome's built-in AI—no external servers or API keys required
- Comprehensive Analysis: Successfully integrated privacy policy parsing, cookie inspection, AI summarization, and risk scoring into a cohesive system
- User-Friendly Design: Transformed complex legal documents into simple risk scores and actionable recommendations anyone can understand
- Fallback Resilience: Built robust heuristic-based fallback analysis for environments where AI APIs aren't available
- Privacy-First Architecture: Created a privacy tool that itself respects user privacy—no data collection, no tracking, no network requests (except fetching same-origin policies)
What we learned
- Chrome AI Capabilities: Deep insights into Chrome's on-device AI APIs and their potential for privacy-preserving applications
- Privacy Policy Patterns: Common structures and red flags in privacy policies, including data collection practices and deletion rights language
- Extension Architecture: Best practices for Manifest V3, service workers, content script communication, and local storage management
- AI Prompt Engineering: How to craft effective prompts for consistent, structured JSON outputs from language models
- Privacy Frameworks: Understanding of GDPR, CCPA, and other privacy regulations that inform best practices
What's next for Safe Mode
- Enhanced Privacy Framework Support: Add specific indicators for GDPR compliance, CCPA rights, and other regulatory frameworks
- Historical Privacy Tracking: Monitor and alert users when websites change their privacy policies, especially to become more invasive
- Batch Analysis: Enable users to analyze all open tabs or their frequently visited sites at once
- Privacy Score Trends: Show how websites' privacy practices compare to industry averages and track improvements/degradations over time
- Community Privacy Database: Optional opt-in sharing of anonymized risk scores to build a crowd-sourced privacy safety database
- Browser Support Expansion: Extend to Firefox and Edge once they support similar on-device AI capabilities
This submission highlights the innovation, technical complexity, and user value of your Safemode project while telling a compelling story about why it matters!
Log in or sign up for Devpost to join the conversation.