Pope Leo XIV Just Wrote a 235-Page Encyclical on AI. Anthropic's Co-Founder Was Standing Next to Him.

The Holy See released Pope Leo XIV's first encyclical at the Vatican this morning. 235 pages, 245 paragraphs, five chapters, the whole thing dedicated to one subject: what artificial intelligence does to the human person. Magnifica Humanitas is the longest opening encyclical of any modern pope, and it is the first to take AI as its central question.

That is the part the headlines will lead with. The part the headlines will under-read is who was standing on the dais next to him.

Pope Leo XIV personally presented the document. No pope has done that before. The job has always gone to a cardinal or a senior curial figure. Today the pope took the lectern himself, and the layperson at his side was Chris Olah, co-founder of Anthropic. Not a head of state. Not a UN secretary general. The interpretability researcher from a frontier AI lab that is, as of this week, closing a $30 billion round at a $900 billion valuation.

You can read the encyclical for its theology. You can read it for its policy. I want to read it for its staging, because the staging is doing political work that the text alone cannot.

What Is Actually In Magnifica Humanitas

The encyclical was signed on May 15, 2026. That date is not incidental. It is the 135th anniversary of Rerum Novarum, the 1891 Leo XIII text that defined Catholic social doctrine for the first Industrial Revolution and set the frame on labor, capital, wages, unions, and the obligations of property owners. Picking the name Leo was already a signal. Signing the AI encyclical on the Rerum Novarum anniversary is the same signal, twice as loud. He is telling the Church and everyone reading along that AI is the new industrial question, and that he intends to do for it what Leo XIII did for factory labor.

The text covers a lot. Five chapters move from a historical arc of Catholic social teaching, to the human-dignity foundations, to labor, to truth, to peace and warfare. The arguments that will land in policy circles are concentrated in the back half:

On autonomous weapons. The pope writes that AI use in war must be subject to "the most rigorous ethical constraints" and that it is "not permissible" to entrust AI systems with lethal decisions. He goes further than that. He says some autonomous weapons systems have advanced "practically beyond any human reach to govern them." That is a Holy See document, signed by the pope, asserting that humans have lost meaningful control over deployed weapons systems. The phrasing is going to be cited in every arms-control hearing for the next decade.

On data. He calls for "wider ownership of AI data" as a matter of justice, framing concentrated training data the same way Leo XIII framed concentrated capital. This is the encyclical quietly endorsing a distributive politics of model inputs without using the word distributive.

On labor. Workers' rights need protection in the AI transition, and the protection has to be active, not hopeful. The text rejects the technological-utopia framing and rejects the civilizational-collapse framing in the same breath, asking the reader to hold both possibilities and act as if both could happen.

On children. Keep them away from the surfaces that are not safe for them. This is the encyclical aligning with the GUARD Act's framing in the United States and the EU AI Act's framing in Brussels at the same time.

On governance. Slow down. Regulate closer. Set shared standards of social justice and apply them before deployment, not after harm.

Why Chris Olah, Specifically

The Vatican does not pick stage partners by accident. Choosing Chris Olah, and not for example Sam Altman or Demis Hassabis or Dario Amodei or Elon Musk, says something. Olah is the interpretability researcher. He is the public face of the "we need to understand what the models are doing internally" school of AI safety inside the lab that has made interpretability and constitutional AI its main marketing surface.

Anthropic has spent two years framing itself as the safety-first frontier lab. The framing is not free, it costs them compute and it costs them deployment speed, and we have written before about the ways that bet has played out commercially. Today the bet got an endorsement that no PR budget can buy: the universal Catholic Church signaled, by who it put on the lectern, that this lab's framing of the AI safety question is the framing it finds serious.

That is not the same as the pope endorsing Anthropic. He did not. The encyclical does not name any company. But the visual is the visual. A frontier lab co-founder in the same camera frame as the pope, on the day the church takes its formal position on AI, is a form of moral underwriting that will follow Anthropic through every regulator meeting and every enterprise procurement conversation for years.

The Week Lined Up Too Cleanly to Be Random

Look at what landed in the same five business days.

OpenAI's answer to scale is to take the company public, with Goldman and Morgan Stanley leading, at a valuation somewhere between $852 billion and $1 trillion. The frame is American capital markets. The argument is "we are an institution now, the public should own a piece."

Anthropic's answer is different. Same week, comparable amount of capital, but raised privately, at a valuation that surpassed OpenAI's for the first time. And then on the same Monday, their co-founder is at the Vatican with the pope. The frame is legitimacy through gravitas rather than legitimacy through float.

I do not think this is coincidence. The Vatican prepared the encyclical for months. Anthropic prepared its round close for months. The timing converges because both organizations have a shared interest in the answer to one question: what is the legitimate frame for frontier AI deployment? The pope answered it today. He framed AI as a moral question on the order of factory labor in 1891. Anthropic is the lab that has been arguing the same thing.

What Moral Capital Actually Buys

People reading this who work in tech will roll their eyes. An encyclical does not change a benchmark, does not ship a model, does not lower a price. Fair. So let me be specific about what it does do.

It hands regulators a vocabulary. Until today, every AI safety hearing was conducted in language the labs themselves invented: alignment, RSP, frontier model, capability evaluation. Now there is a 235-page document, signed by a moral authority older than the United States, that imports the language of Catholic social doctrine onto the same problem. Senators will quote it. Health ministers will quote it. The EU rapporteurs already drafting the next round of the AI Act will quote it. Vocabulary is not nothing. The labor movement of the twentieth century ran on Rerum Novarum vocabulary for decades.

It binds the conversation about autonomous weapons. The pope did not just call them dangerous, he said they were already beyond governance. The Pentagon, which signed classified-network AI deals with OpenAI, Google, Microsoft, AWS, NVIDIA, SpaceX, and Reflection earlier this month (and pointedly not with Anthropic, the only frontier lab with a public no-weapons policy), now has to argue with a pope on top of arguing with arms-control NGOs. The shape of that argument is structurally harder.

It establishes a comparison standard for safety claims. Every lab that pitches its safety story now has to answer the interpretability-and-constitution framing that Olah and Anthropic spent years building. The pope's presence does not validate the technical work, but it does validate the seriousness of the framing, and the framing was always going to win or lose on seriousness, not on benchmarks.

It changes the enterprise sales motion. Catholic hospitals. Catholic universities. Catholic banks. Catholic charities. Roughly 1.4 billion baptized Catholics globally, and the institutions that spend on their behalf are large, conservative, and risk-averse on ethics. Procurement at a Catholic health system that just got 245 paragraphs of guidance is going to look very different from procurement at one that did not.

The Counter-Read

The honest counter goes like this. Encyclicals are non-binding, frontier capability is set by people who have not heard of Rerum Novarum, and AI deployment is going to happen on the same Q3 ship cycles whether the Vatican signs a document or not. The labs that actually decide what humans build with AI are American and Chinese, and the American ones answer to the SEC and DARPA before they answer to a pope. China answered no one at all on this question. The encyclical is words.

That read is not wrong, it is just early. Rerum Novarum was words too, in 1891. It did not stop the steel mills or unionize the docks. What it did was sit in the background of every labor fight for the next century and give the people doing the fighting a vocabulary that could not be dismissed as merely economic. It made the moral cost of certain choices legible to people who otherwise had no language for them. Magnifica Humanitas is reaching for the same role. It will succeed at the same delay.

Three Signposts To Watch

First, citation patterns. Watch which AI policy documents, hearings, and regulatory proposals start citing Magnifica Humanitas inside the next ninety days. Rerum Novarum took roughly five years to become a standard reference. The information cycle is faster now. If we see Senate committee citations and EU rapporteur citations inside three months, the encyclical is functioning as policy infrastructure rather than as theology.

Second, lab response patterns. Watch which frontier labs publicly engage the text and which do not. The labs that ignore it are telling you they think the moral frame is not going to bind them commercially. The labs that engage it, especially the ones that quote it favorably, are telling you they have decided moral capital is worth more than the speed cost of acknowledging the constraint. Anthropic has already shown its hand by sending Olah to the presentation. The others have a choice to make.

Third, Pentagon procurement. Anthropic is out of the May 1 classified-network deals because of its no-weapons policy. After today, the pope is on the record arguing that lethal AI decisions are "not permissible." If Anthropic finds itself back at the table for the next round, or if a parallel non-weapons defense procurement track opens up, you will know that the moral framing is translating into procurement reality.

The Read

A pope wrote 235 pages on AI and presented them himself, breaking a centuries-old protocol, with a frontier-lab co-founder at his side on the same day his rival filed for the largest tech IPO in U.S. history and his preferred lab closed the largest private AI round on record. That is not three news items. That is one news item, which is that the AI industry is now negotiating with three authorities at once: capital markets, governments, and the Holy See. Until today, only two of those were in the room.

We are going to be quoting Magnifica Humanitas on this site for the next several years. The model wars are still real. The benchmark fights are still real. But the layer above them, the layer where deployments live or die because of what the broader public thinks they mean, just got a new floor.

The encyclical full text is at the Vatican website. The Anthropic round is expected to close formally this week. You can see the frontier lab valuations we track on our funding page and the model catalog on our models page. Watch the citations.