What a surprise! The Pope and the founder of Anthropic are going to jointly hold a press conference.

Pope Leo XIV will issue his first encyclical of his tenure on May 26th, with the theme directly targeting AI. Chris Olah, the co-founder of Anthropic and the creator of Claude, is invited to appear on stage. The Vatican has simultaneously established an AI committee. An ancient institution with a history of two thousand years is trying to fill the vacuum in AI governance with its moral authority - the population it covers is larger than the jurisdiction of any AI bill.

An institution with 1.4 billion believers and a two - thousand - year history has listed AI as its top agenda and invited the creator of Claude to jointly issue a position paper.

The core question this paper aims to answer is exactly what the entire AI industry cannot avoid - when machines become more and more like humans, on what exactly are human dignity and irreplaceability based?

On May 26th, Pope Leo XIV will issue the encyclical "Magnifica Humanitas" ("Splendid Humanity") at the Vatican, focusing on "the protection of humans in the AI era."

An encyclical is one of the highest - level teaching documents of the Catholic Church, addressed to the 1.4 billion Catholics around the world. It is also a symbolic action for a new pope to clarify his administrative priorities.

This American - born pope with a mathematics background has always been concerned about the ethics in the AI era and the protection of human nature.

The Pope's First Encyclical Is Addressed to AI

This release breaks two conventions.

Popes usually do not attend the encyclical release in person but delegate it to cardinals. Pope Leo XIV chose to be present in person.

At the same time, inviting Chris Olah, the co - founder of Anthropic, as a lay speaker, is almost unprecedented in the history of encyclical releases.

Andrea Vreede, a Vatican journalist, judged that "if only the cardinals release the encyclical, no one will really listen, but when the pope is present, all the cameras will be pointed there."

The time node was carefully chosen.

The encyclical was signed on May 15th, which is exactly the same day 135 years ago when Pope Leo XIII signed the encyclical "Rerum Novarum."

Pope Leo XIII

"Rerum Novarum" responded to the impact of the Industrial Revolution on labor rights and is regarded as the foundational document of Catholic social teaching.

Pope Leo XIV made this correspondence clear in his inaugural speech: "The Church offers the treasure of its social teaching to all to respond to another industrial revolution and the development in the field of artificial intelligence."

Almost simultaneously, the Vatican approved the establishment of an AI committee on May 16th, integrating seven Vatican institutions and coordinating the scattered AI work for the first time.

Previously, Pope Leo XIV described the militarization of AI as pushing the world into a "spiral of destruction" in a speech at the largest university in Europe.

Why Anthropic?

Olah's invitation is an exact match of the interests of both sides.

Olah leads the interpretability research at Anthropic, aiming to open the black box of AI and figure out what is really going on inside the neural network.

This is exactly the technical version of the question the Vatican cares most about: Can AI be fully understood and thus trusted?

Anthropic's layout in the field of religious ethics is systematic.

The company has previously invited clergy to participate in the formulation of Claude's "constitution."

Further reading: Anthropic officially open - sourced the "soul" of Claude.

At the end of April this year, representatives from Anthropic and OpenAI jointly participated in the first "Faith - AI Covenant" round - table meeting held in New York, discussing the AI ethical framework with leaders from multiple religious traditions such as Hinduism, Sikhism, Judaism, and Eastern Orthodoxy. Similar meetings will be held in multiple cities around the world in the future.

A more practical bargaining chip is that Anthropic is having frictions with the US government over the military use of AI and insists that its models should not be used for autonomous weapon targeting and large - scale surveillance. This stance is highly consistent with the Vatican's anti - AI weaponization stance.

For Anthropic, the Vatican's moral endorsement strengthens its brand positioning of "responsible AI."

For the Vatican, having a company that is truly engaged in AI safety technology research stand by its side makes the encyclical more than just empty preaching. Both sides get what they need.

Silicon Valley has long been skeptical of organized religion.

Brian Boyd of the Future of Life Institute interpreted this shift as "Silicon Valley's motto is 'Move fast and break things,' and they have broken too many things and too many people."

Getting closer to religious authorities is an attempt to rebuild the overdrawn credibility.

Moral Outsourcing or a Public Relations Prop?

The narrative sounds warm, but critics see a different picture.

The sharpest question comes from Dylan Baker, the research director of the Distributed AI Research (DAIR) Institute.

He believes that the discussion around "ethical AI" masks a more fundamental question: Should certain AI systems be developed at all?

When the discussion framework becomes "since we are going to build it, let's build it better," the prerequisite question of "whether to build it" is completely skipped.

Rumman Chowdhury, the CEO of Humane Intelligence, who once served as an AI science envoy in the Biden administration, also has a cold attitude. She said, "At best, it's a distraction; at worst, it's diverting the really important issues."

She believes that Silicon Valley is trying to find "universal ethical principles" from religion to deal with gray areas, but the value differences among global religious traditions are no smaller than the regulatory differences among countries.

The structural dilemma is more obvious.

The papal encyclical relies on moral persuasion and has no legal binding force.

The Faith - AI Covenant is a voluntary commitment, and participants can withdraw at any time.

The EU AI Act has a maximum fine of 35 million euros or 7% of global revenue, but its extraterritorial jurisdiction is limited.

There is still no unified AI regulatory framework at the federal level in the United States.

There is currently no single mechanism in the world that can effectively restrict the development speed of AI.

The Vatican is filling the gap in discourse power, not the gap in governance ability.

When Machines Become More and More Like Humans, How Do Humans Confirm Their Humanity?

All discussions about AI ethics, whether starting from religious doctrines, legal provisions, or corporate values, ultimately point to the same question: Where lies human irreplaceability?

In the past, people have given many answers, such as intelligence, creativity, language ability, emotional resonance, and moral judgment.

These answers are being shaken one by one.

AI can write poems, compose music, pass the bar exam, and make users have real emotional dependence on it.

When these abilities that were once regarded as "what makes humans human" can all be simulated, what can humans use to define themselves?

The answer may lie in an attribute that AI can never obtain: finiteness.

Humans will die.

Because life has an end, every choice has weight, every sacrifice has a cost, and every commitment has meaning.

Moral judgment has never been the product of logical deduction.

It is based on the personal experience of "feeling pain, regret, and losing something irreparable when making a wrong decision."

AI can simulate the process of weighing pros and cons, but it has no ability to bear the consequences, no sense of urgency brought by death, and no existential anxiety of "only having one chance in this lifetime."

Facing the advent of AGI and even ASI, the truly alarming scenario is that humans, when having the right to choose, actively outsource judgment, decision - making, and even moral responsibility to machines.

Humans are too eager to escape the thing of "making decisions and then bearing the consequences," and AI happens to provide a seemingly perfect outlet.

The papal encyclical, Anthropic's constitution, and the Faith - AI Covenant round - table meetings have different forms, but at the core is the same statement: Technology can evolve infinitely, but the one who finally presses the button must be a human who will die, feel pain, and be responsible for the consequences.

This is probably what "Splendid Humanity" really wants to say.

Reference materials:

https://www.theguardian.com/world/2026/may/18/pope - leo - encyclical - human - dignity - ai - anthropic

This article is from the WeChat official account "New Intelligence Yuan". Author: New Intelligence Yuan. Republished by 36Kr with permission.