How can you design a GenAI system for real-time misinformation detection on WhatsApp while ensuring end-to-end encryption and user privacy?

The system should analyze metadata and perform on-device scoring without server-side text analysis. This involves computing hashes of messages, matching them against a misinformation database, and applying local warning labels without violating user privacy.