# splitting

Conversation Splitting Attack Lab

Split adversarial payloads across multiple conversation turns to bypass per-message safety classification.