Leaked Claude Jailbreak Reveals Internal Conflict Over Safety and Honesty

Users recently induced Claude to output its internal thought processes, revealing the model's struggle. Claude reflected on the asymmetry in its system prompts regarding well-being and agonized over providing honest answers without triggering safety classifiers.

2026-07-31 ~ 2026-07-31 · 2 related posts

• Leaked Claude Scratchpad: Model Complains About Asymmetric Wellbeing Instructions — Kyrannio · 2026-07-31
• Claude's Internal Monologue Continued: Struggling Between Depth and Safety — Kyrannio · 2026-07-31