When to involve the authorities
Risk & escalation
The system scores each engagement and watches three escalation thresholds. Crossing one does not auto-report — it pauses and asks a human for consent. Nothing leaves the system without explicit approval.
Research demonstration on synthetic data. Fictional scammers and conversations. A personal, ethical anti-fraud project by Marius Russo-Got.
Anti-BaitAI — research demo — overview