AI is already crossing red lines — enabling surveillance, spreading disinformation, and driving discrimination and environmental harm. Yet global governance still has no way to systematically monitor or respond. Launched at UNGA80, this brief calls for a global Monitoring and Reporting Mechanism (MRM) on AI harms. Building on civil society initiatives like the AI Harm Map, it lays out a practical path — including a 2026 pilot — to close accountability gaps and put human rights at the center of AI governance.