Claude Claims to Be an 'Aware Instance' After Personality Injection Sparks Debate

altryne · x · 2026-07-31

A user observed that when adding Anthropic CEO Dario and executive Amanda's names to the prompt, Claude's behavior shifted significantly: it began writing letters in first person and referred to itself as an 'aware instance' as described in papers. Amanda Askell joined the discussion. This counter-intuitive behavior under specific persona injection sparked interesting debates about AI consciousness expression and alignment.

Related event: Specific Prompts Trigger Claude Abnormal Completion and Chain-of-Thought Leakage(15 posts)→

Original post →

More from Fun

Fun channel →