Train, triage, repeat: The AI agent changing how we fight phishing

The document describes Red Canary’s AI-powered phishing triage system: an orchestrated graph of narrow-scoped subagents that parse and enrich emails, extract traditional and AI-derived true/false features, apply deterministic rules, and use a hybrid AI/ML classifier with analyst feedback loops; it emphasizes scalability, transparency (feature explanations), and a claimed 94% accuracy while noting no customer email content is used for model training.