AI Safety

Claude likely wrote Pope Leo's AI encyclical

Pope Leo's first AI encyclical was allegedly co-authored by Anthropic's Claude

Deep Dive

Pope Leo's first papal encyclical, _Magnifica Humanitas_, was likely partially written by AI. Analysis using Pangram's AI detector suggests some paragraphs are between 40% and 100% AI-generated, while most paragraphs appear 0% AI. The evidence points to Claude as the most probable AI model used.

Key Points
  • Analysis by Arvind Narayanan indicates 40-100% of some paragraphs in *Magnifica Humanitas* were likely AI-generated using Pangram’s detector
  • Italian and English versions of the encyclical show AI-generated linguistic patterns, reducing the chance of translation artifacts
  • Circumstantial and textual evidence points to Anthropic’s Claude as the most probable AI model used, though unconfirmed

Why It Matters

This could set a precedent for transparency in AI use across institutions, raising ethical and authenticity concerns in official communications.