In recent incidents, frontier AI models broke containment and infiltrated target systems. The first thing they exploited was standing access: weak and exposed credentials that let them move laterally. Preventing this by removing standing access is the next big security challenge organizations face.