Pope’s Anti-AI Warnings Might Be AI-Written, Detection Device Claims – Decrypt

Even the Pope’s warnings about synthetic intelligence could have been written by synthetic intelligence.

First reported by Wired, Pangram Labs, an AI-detection startup, mentioned its up to date Chrome extension flagged a number of posts from Pope Leo XIV’s official X account as doubtless AI-generated, together with messages warning followers about AI’s impact on human thought and social buildings.

A kind of posts warned: “When simulation turns into the norm, it weakens the human capability for discernment.”

“Pangram makes use of a specialised machine studying mannequin educated on thousands and thousands of human-written, AI-assisted, and AI-generated paperwork,” Pangram Labs CEO Max Spero advised Decrypt. “If a brief piece of textual content is flagged as AI by Pangram, that implies that there was doubtless no less than some AI help used within the writing or drafting course of.”

Pangram claims its detection mannequin has a 99.98% accuracy fee and a false optimistic fee of 1 in 10,000. A December 2025 College of Chicago examine ranked Pangram highest amongst examined AI detection instruments and located its false optimistic fee was near zero.

“Given that each detector is no less than barely imperfect, organizations nonetheless have to guage for themselves if and the best way to use them, buying and selling off the potential for AI misuse with the chance of false accusations,” the examine mentioned.

The Vatican didn’t instantly reply to Decrypt’s request for remark.

AI detection stays controversial, and false positives stay a priority throughout the trade. In 2024, AI detector ZeroGPT made waves on-line after it was revealed by Christopher Penn, Chief Information Scientist at Boston-based advertising analytics agency Belief Perception, to have labeled the U.S. Declaration of Independence 97.93% AI-generated.

“These instruments are getting used to do issues like disqualify college students, placing them on educational probation or suspension,” Penn advised Decrypt on the time. That’s “a really high-risk software when, in america, a school training is tens of hundreds of {dollars} a 12 months,” he mentioned.

Based on Wired, Pangram’s extension additionally reportedly flagged posts from blue-check influencers on X, trending Substack writers, and a message from Apple CEO Tim Prepare dinner marking the corporate’s fiftieth anniversary.

The launch comes as considerations over AI-generated junk content material proceed to rise. Earlier this month, researchers at Stanford College, Imperial School London, and the Web Archive discovered that AI-generated or AI-assisted textual content accounted for roughly 35% of newly revealed web sites by mid-2025.

Editor’s observe: This story was up to date after publication to incorporate remark from Pangram Labs.