The Vatican entered into a dispute with AI detectors over the first encyclical of Leo XIV - hi-Tech.ua

The Vatican Just Showed Why One AI Detector's Verdict Isn't Enough

Pope Leo XIV's first encyclical, Magnifica Humanitas, addresses human dignity in the age of artificial intelligence — and promptly became a test case for AI detection itself. According to hi-Tech.ua, reporting on 24 July 2026, the detection service Pangram Labs flagged several posts on the Pope's official account, and later parts of the encyclical, as likely machine-generated. Researcher Lynch Zhang offered a similar reading, suggesting any AI involvement was more likely Vatican staff drafting aids than the Pope's own writing. Months later, the Holy See commissioned a second firm, Proudly Human, which compared the encyclical's style against past papal speeches and addresses and concluded that both the encyclical and the earlier Maps of Hope collection were fully human-written.

Did two detectors really reach different verdicts on the same text?

Yes — Pangram Labs flagged sections as likely AI-written, while Proudly Human later certified the identical encyclical as fully human-made.

Neither company has published a side-by-side breakdown of which passages triggered which score, or why. That gap matters more than the headline disagreement: a reader has no way to check whether the two tools were even scoring the same fragments, or whether one caught a stylistic quirk the other's model wasn't trained to see.

How can a "human authorship" check itself rely on AI?

Proudly Human runs its stylometric and grammatical analysis through several specialized AI models, then combines their outputs into one authorship score.

That's not a contradiction, exactly — using models to analyze style is different from using them to generate text. But it does mean the "independent" clearance the Vatican received was itself a machine judgment, layered on top of another machine's earlier flag. The certification is really a claim about a proprietary algorithm's confidence, not a hand-verified fact.

What should buyers ask before trusting a clean detector result?

Ask which model versions generated the score, how "human-made" is defined, and whether a second vendor's tool agrees before relying on either verdict.

Vendors in this space rarely publish comparable methodology, and the encyclical episode shows a single scan — clean or flagged — can swing a document's public reception. Anyone using a detector to clear or condemn a piece of writing, whether for a newsroom, a school, or a publisher's AI policy, is buying a probability estimate from one company's black box, not a fact.

Does this change how a single flag should be treated going forward?

Treat any one detector's score as a starting point for review, not a final ruling, particularly when reputational or academic stakes are high.

Vendors routinely describe false positives as rare edge cases. The Vatican dispute is a reminder that "rare" still means real documents get flagged, and that resolving a flag currently means paying for a second, differently-branded opinion rather than a transparent re-check of the same evidence.

Frequently asked questions

What is Magnifica Humanitas?

It's Pope Leo XIV's first encyclical, published in 2026, addressing human dignity as artificial intelligence becomes more widespread.

Who is Proudly Human?

A company hired by the Vatican to certify that texts were produced without generative AI, after Pangram Labs' initial flags on Vatican materials.

Did the Vatican confirm AI was used anywhere in the document?

No. Proudly Human's certification states the encyclical and the earlier Maps of Hope collection were entirely human-authored, with no AI-written sentences or paragraphs.

Sources: hi-Tech.ua.