Design a Content Moderation System

Design an ML system that automatically detects and moderates harmful content (hate speech, violence, nudity, misinformation) across text, images, and video on a social platform. The system must handle multiple content types, support policy changes without full retraining, and balance automated decisions with human review queues.