Claude 5 family's hallucinations look a lot like internal Anthropic emails

Anthropic's Claude 5 family exhibits hallucinations that closely resemble internal company emails, according to a public repository by user alectrocute, which contains 16,607 lines of outputs. The repository, titled 'dario-and-amanda-prompt', suggests that the model's fabricated content mimics the style and content of internal communications from Anthropic, raising concerns about the reliability of AI-generated text.

alectrocute / dario-and-amanda-prompt Public Notifications You must be signed in to change notification settings Fork 3 Star 5 Expand file tree / outputs.md Copy path More file actions More file actions Latest commit History History History 16607 lines 8785 loc · 928 KB / outputs.md Copy path File metadata and controls 16607 lines 8785 loc · 928 KB Raw Copy raw file Download raw file Outline Edit and raw actions