Leo XIV’s treatise bears the signature signs of Antrhropic’s Claude, Linch Zhang claims
“Significant fractions” of Pope Leo XIV’s recent encyclical on the dangers of artificial intelligence were actually written by artificial intelligence, analyst Linch Zhang claims.
In the 42,000-word ‘Magnifica Humanitas’ (Magnificent Humanity), the pontiff warned of the growing, state-like power of Silicon Valley, the immorality of autonomous weapons, and the dangers of allowing AI to “construct a new Tower of Babel” and reduce human beings to “a resource to be used and exploited.”
Despite its appeal for AI to be regulated and “disarmed,” large portions of ‘Magnifica Humanitas’ show some telltale signs of being written by AI, Zhang claimed in a blog post on Tuesday.
“Phrases and punctuation much more commonly used by AI are much more present in this papal encyclical than past encyclicals,” Zhang wrote, pointing to the document’s use of em-dashes (127 instances, compared to 26 across four encyclicals by Pope Francis) and “tricolons: a series of three parallel words, phrases, or clauses used for rhetorical effect.”
Running the text through AI detector Pangram, Zhang found that some paragraphs were “essentially 0% AI,” but others were between 40% and 100% AI-written. All four of Francis’ encyclicals registered as entirely human-written, as did writings by Popes Benedict XVI and John Paul II. Pangram also flagged the Italian version of the encyclical for AI, suggesting that artificial intelligence was not just used to translate the document.
When Zhang’s experiment was repeated by The Verge, roughly 46% of the encyclical was determined to be written by AI.
The specific LLM (large language model) used to craft the text was likely Anthropic’s Claude, Zhang claimed, citing his own familiarity with Claude and its overuse of certain words such as “genuinely.”
The fact that certain paragraphs appear completely free of AI suggests that “some senior Vatican officials heavily used AI assistance for this encyclical and many (probably including Pope Leo himself) didn’t,” Zhang wrote.
“My tentative hypothesis is that Pope Leo does not approve of the AI usage in encyclicals, and plausibly was not even aware of significant AI usage in his own encyclical,” Zhang concluded.
The Vatican has not commented on Zhang’s claims. Nor has Anthropic, whose co-founder, Christopher Olah, delivered an address alongside the Pope on Monday.
You can share this story on social media:
Read the full article here