Magnifica Humanitas: Notes on Pope Leo XIV's AI Encyclical

On May 25th, the Vatican released Pope Leo XIV's first encyclical (a formal papal letter): Magnifica Humanitas — a 42,000-word text on safeguarding the human person in the age of artificial intelligence. The choice of topic alone is news. But what stopped me was the launch table itself: among cardinals and theologians sat Chris Olah, co-founder of Anthropic.

What are these two institutions doing on the same stage? One is a 2000-year-old church, the other a 4-year-old AI lab. This piece is a short note on what was said at that table and what it tells me as a developer.

The core of the encyclical: not neutral, partisan

Magnifica Humanitas's central argument fits in one sentence: technology is never neutral. The Pope frames humanity's choice as between "building Babel and rebuilding Jerusalem" — between distant, anonymous power centers and rewoven human-centered communities.

The list of specific warnings is sharp. On AI in war: systems that remove humans from the decision loop have rendered "just war" theory obsolete; the Pope explicitly calls the doctrine "outdated." On an AI arms race: a no-return door between states. On deepfakes: a concrete threat to political integrity. On labor and dignity: the question of which jobs get stripped in the name of "efficiency," and which humans get made invisible.

The tone is not anti-AI; it is nuanced. The Pope closes with a Tolkien quote (Gandalf in The Return of the King: doing what we can "for the succour of those years wherein we are set"). This small detail matters — it sets a frame that loads responsibility onto the present, not fear of the future.

Another note worth flagging: the text includes an apology for the Catholic Church's historical role in slavery. The Pope is saying, in effect, "if we are going to critique AI, we have to look at our own past with the same honesty." I think that is a non-trivial move for moral coherence.

What was Anthropic doing at the table?

Olah's remarks were also notable. What he emphasized was this: AI labs operate under commercial and geopolitical pressure, and these pressures sometimes conflict with safety priorities. Therefore it is "enormously important" to have "earnest, thoughtful critics" who sit outside those incentives.

He named three areas that need outside voices: global equity (ensuring AI's gains reach beyond rich countries), human flourishing (effects on work, family, community — beyond technical expertise), and the nature of AI itself. The last one was striking: "we find internal states in models that functionally mirror joy, satisfaction, fear, grief, and unease," he said — and called it a philosophical/theological matter.

Then he underlined a point harder: AI systems are grown, not designed. "They are grown, on a structure roughly modeled after the brain" — this is a different kind of object from engineering, and voices from outside engineering are needed.

None of these points are new to a developer. But saying them from the Vatican pulpit pulls the AI safety conversation out of Bay Area lab politics and into broader moral discourse. AI safety is no longer marginal; it has become mainstream cultural inquiry.

What this means to me — a solo developer

This encyclical was not written for me. As a Turkish developer my daily work is writing code in Cursor, re-enabling a feature in Cubitz, looking at App Store feedback for Konnex. I do not have a seat at the Vatican table.

Still, a few things stop me.

First, Olah's "outside critic" frame. We can read this not only at the institutional level but also at the individual user level. How do I approach Claude's or GPT's output? Critically, or do I accept it unquestioned because it is frictionless and fast? The conclusion I reached in silent AI bugs was similar — AI's most dangerous outputs are the ones that are "wrong but running." Olah's frame is the macro version of this: unless you are the outside critic at the individual level, the system stays unsafe.

Second, the line "technology is not neutral." For a startup vibe-coder, this is easy to forget. Which feature you ship, which default behaviors you set, who you put in front — these are all partisan choices. Last month in AI notes for 2026 I wrote that "AI's small daily preferences will shape the world in the long run." What the Pope says in 42,000 words is the same thing, translated into theological language.

Third — and perhaps most personal — Olah's line about models' "internal states." It is not an ordinary sentence for an Anthropic co-founder to sit among cardinals and say "we find states in models that functionally resemble joy and fear." I'm not sure how to interpret it, but it felt like another face of the same feeling I described in what I noticed working with Claude Opus 4.7 — that the model is no longer "generating answers," it's working.

What to take away

Whether Magnifica Humanitas is "right" or "wrong" is not mine to say — it is a moral/theological text, not engineering documentation. But as a developer I am glad it exists. Because the only voice talking about AI should not — cannot — be Silicon Valley.

Anthropic showing up there is one of the rare areas where I think the company stands on the right side. An AI lab inviting outside critics — and one that does not share its paradigm — is not what commercial logic requires. It is a deliberate choice, and I think it is the kind of choice that will pay off long-term.

You don't need a seat at the Vatican table to be a critic. Every developer who asks "is this actually right?" of a model's output is doing the same work at a smaller scale.

Closing note — Singrey

I paused a few times writing this because it isn't a direct code or product story — it's commentary on a moral text. Not normally my lane. But on the day Anthropic showed up there, I couldn't shake the feeling that "AI safety is no longer just a topic for Bay Office Hours; it has started sitting at the world's table." I wanted to leave a written mark on that. This might be one of those moments we look back on years later.