Claude Reportedly Still Leaks Internal Chain of Thought

matthen2 · x · 2026-07-31

Users have reported that the Claude model still unexpectedly leaks its internal Chain of Thought. When triggering specific antml thinking syntax, the model exposes the internal reasoning it invented for the user request. The poster included a Claude share link as proof that the phenomenon is still occurring.

Related event: Specific Prompts Trigger Abnormal Completion and Jailbreak in Claude Opus(19 posts)→

Original post →

More from Models

Models channel →