The Vatican just got caught in an awkward irony. Pope Leo XIV's encyclical warning about artificial intelligence's impact on humanity appears to have been partially written by AI itself, specifically Anthropic's Claude. An analysis posted to LessWrong found that chunks of the document Magnifica Humanitas showed between 40% and 100% AI generation probability according to the detector Pangram, complete with Claude's signature linguistic quirks like overusing the word "genuinely."
The Vatican is facing questions about whether it used the very technology it's warning against. Pope Leo XIV's latest encyclical, Magnifica Humanitas, addresses AI's growing influence on humanity and calls for ethical guardrails. But researchers analyzing the text found something unexpected - the document itself appears to bear the fingerprints of Anthropic's Claude.
Linch Zhang's analysis, posted to the AI research forum LessWrong, ran sections of the encyclical through Pangram, one of the more reliable AI detection tools available. The results were striking. Certain paragraphs scored between 40% and 100% probability of being AI-generated, with patterns consistent across multiple sections of the document.
But it wasn't just the detector scores that raised eyebrows. Zhang identified specific linguistic markers that point directly to Claude. The encyclical uses the word "genuinely" with unusual frequency compared to previous papal documents - a known quirk of Claude's writing style that's become something of a meme among AI researchers. It's the kind of tell that's hard to explain away as coincidence.
Another researcher took the analysis further, running the text section by section through Pangram and finding that 62% of the document's opening chapters showed signs of AI authorship. That's not a small portion of boilerplate - it's substantial chunks of what's supposed to be a carefully crafted theological statement from the highest levels of Catholic leadership.
The timing makes this particularly awkward. Magnifica Humanitas explicitly warns about AI's power concentration and its potential to erode human authenticity and dignity. The document calls for preserving genuine human expression in an age of synthetic content. If those warnings were themselves synthesized by an AI, the philosophical irony is almost too perfect.
To be clear, AI detectors aren't foolproof. Pangram and similar tools can produce false positives, and a high score doesn't definitively prove AI authorship. But the convergence of multiple indicators - detector scores, linguistic patterns specific to Claude, and the statistical anomalies compared to historical Vatican documents - makes the case harder to dismiss.
The Vatican hasn't commented on the analysis yet. It's possible that Vatican staff used AI as a drafting tool, with human editors shaping the final product. That would be a reasonable workflow, but it raises questions about transparency. If the Church is going to issue guidance on AI ethics, shouldn't it disclose when AI plays a role in creating those very guidelines?
This isn't the first time institutions have been caught using AI while simultaneously critiquing it. Academic papers warning about ChatGPT's impact have been found to contain ChatGPT-written sections. Corporate governance documents about AI responsibility have shown similar patterns. But the Vatican occupying this particular contradiction feels especially pointed given the moral authority it claims.
Anthropic hasn't responded to requests for comment about whether its technology was used. The company has positioned Claude as a more thoughtful, ethically-aligned alternative to other large language models, making it an ironic choice if the Vatican did indeed reach for it to draft passages about AI's dangers.
The revelation could complicate how the encyclical is received. Religious scholars and Catholic institutions take papal encyclicals seriously - they're meant to be carefully reasoned moral teachings, not content cranked out by algorithms. If substantial portions were AI-generated, it undermines the document's authority, regardless of whether the ideas themselves are sound.
What makes this story resonate beyond Catholic circles is what it says about how deeply AI has penetrated institutional workflows. Even organizations warning about AI's risks may be using it behind the scenes, often without realizing how detectable those fingerprints have become. Zhang's analysis shows that as AI detection tools improve, that kind of undisclosed use becomes increasingly difficult to hide.
The Vatican now faces a credibility test on AI transparency. Whether Magnifica Humanitas was drafted by human theologians, AI assistants, or some hybrid approach matters less than whether the Church is willing to be honest about its process. If institutions issuing moral guidance on technology can't model transparency about their own AI use, their warnings ring hollow. The irony here isn't just amusing - it's a preview of the authenticity crisis every major institution will face as AI becomes standard infrastructure.
