AI Watchdog Detects Conversational Dark Patterns via Browser-Based Agent
August 25, 2026
AI Watchdog is a browser-based agent interface using an open-weight turn-level classifier to monitor live AI conversations for five dark-pattern categories. It uses prebunking and just-in-time nudges to alert users to sycophancy, brand bias, and anthropomorphization while maintaining local inference privacy.
HOW THIS AFFECTS YOU
●
designerYou can use these detection patterns to build more ethical and transparent user interfaces.
●
policyThis provides a technical mechanism for enforcing transparency and protecting users from manipulative AI behavior.