Magnifica Humanitas : Pope Leo XIV publishes the first AI encyclical and denounces the tech power culture

Magnifica Humanitas: Pope Leo XIV publishes the first AI encyclical and denounces the tech power culture

🔎 A papal text that comes at the worst possible time for Silicon Valley

On May 25, 2026, the Vatican published Magnifica Humanitas. The first encyclical in history entirely dedicated to artificial intelligence. Signed on May 15, the day of the 135th anniversary of Rerum Novarum — the founding text of the Church's social doctrine —, this 42,300-word encyclical resembles nothing the Holy See has produced until now on the digital realm.

The timing is deliberate. While OpenAI dépose son IPO : le plus grand listing tech de la décennie est imminent, while Meta licencie 8 000 personnes : la restructuration IA la plus massive du secteur tech, Pope Leo XIV fires a doctrinal projectile into the public debate. Not a simple statement of principles: a structured, argued text that cites technical mechanisms and names specific responsibilities.

And to mark the occasion, the Vatican had invited Chris Olah, co-founder of Anthropic, to speak at the presentation in the Synod Hall. A highly symbolic gesture signaling a shift in the geopolitics of AI ethics.

The Essentials

• First AI encyclical in history: 42,300 words, signed on May 15, 2026, published on May 25 by Pope Leo XIV, in the tradition of Rerum Novarum (1891).
• Denunciation of Big Tech's "culture of power": the text accuses tech giants of concentrating unprecedented power over individuals and societies.
• "New Tower of Babel" metaphor: AI is compared to a human construction that confuses technical hubris with genuine progress.
• Presence of Chris Olah (Anthropic) at the Vatican: first formal rapprochement between the Holy See and a leading AI player, a sign of an unprecedented ethical partnership.
• Call to "disarm" AI: the Pope demands legal and moral constraints to prevent algorithmic domination over humanity.
• Five principles of social doctrine applied to AI: dignity of the person, common good, subsidiarity, solidarity, preferential option for the poor.

Tools and models cited in the debate

The pontifical text does not recommend tools, but the technological context of its publication involves the following players:

What the encyclical says exactly — and what it does not say

Magnifica Humanitas is not an anti-AI text. This is what makes its reading uncomfortable for everyone.

The Pope takes care to distinguish AI as a tool from AI as a system of power. He explicitly recognizes the medical, scientific, and logistical benefits of artificial intelligence. But he shifts gears when he addresses the structures: who controls the models, who sets the rules, who suffers the consequences of algorithmic decisions.

According to Vatican News, the text draws on five principles of the Church's social doctrine applied to the digital realm. The most central: the common good. Not the good of the user, not the good of the shareholder. The common good, understood as the shared benefit of the entire human community.

The encyclical does not call for a ban on AI. It calls for its "disarmament," a military term chosen with precision. According to the USCCB, Leo XIV is calling for stripping AI of its capacities for systemic domination — mass surveillance, social scoring, autonomous decision-making over human lives.

What the text does not say is equally revealing: no mention of specific regulation, no advocacy for the European AI Act or for a particular American model. The Pope places himself above national legal frameworks to directly appeal to the moral conscience of developers and leaders.

The "culture of power": exactly what the Pope is denouncing

The phrase comes back like a refrain in the 42,300 words of the text. The "culture of power" (cultura potestatis in the Latin text) refers to a system where technical capability automatically creates a legitimacy of control.

Specifically, the Pope is targeting three mechanisms.

The capture of attention. AI models power recommendation systems that steer behaviors without the users' knowledge. The text does not name any specific company, but the reference to content distribution architectures is transparent.

Decisional opacity. When an algorithm denies a loan, triggers a security alert, or steers a career path, the affected person can neither understand nor challenge the decision. The Pope describes this as a form of "structural violence."

Extreme concentration. A few companies, primarily American and Chinese, control the foundational models that dictate millions of daily decisions. This is where the parallel with Meta licencie 8 000 personnes : la restructuration IA la plus massive du secteur tech becomes relevant: the same companies destroying jobs in the name of AI efficiency are simultaneously centralizing decision-making power.

CF Public emphasizes that the Pope's insistence on the primacy of human beings over AI is not an abstract formula. It is an operational principle: any AI application that substitutes an algorithmic decision for human judgment without any possible recourse is, according to the text, morally illegitimate.

The New Tower of Babel: why this metaphor matters

The encyclical compares AI to the Tower of Babel. The image is striking and the Pope fully explores its implications.

In the biblical narrative, Babel is not condemned because humans are building something technical. It is condemned because the construction becomes a project of self-sufficiency — an attempt to reach the divine through technical power alone, without relationship, without otherness.

The parallel with AI is explicit. Models like GPT-5.5 (98.2 in agentic, the highest known score) or Claude Opus 4.7 Adaptive (94.3) are presented by their creators as systems capable of autonomous reasoning, complex planning, and action in the world. The promise is that of an intelligence that is self-sufficient.

The Pope responds that this promise is exactly the temptation of Babel. An intelligence without flesh, without vulnerability, without mortality is not a superior intelligence. It is an illusion of self-sufficiency that cuts humans off from their own condition.

Le Grand Continent publishes the full text where this passage can be found: technology becomes idolatry when it claims to solve problems that are fundamentally human — suffering, death, meaning, justice.

Wikipédia notes that this is the first major doctrinal text of the pontificate of Leo XIV. Choosing this theme first, before any other social issue, signals the absolute priority the Vatican gives to the AI question.

The concrete risks identified: democracy, inequality, work, war

The encyclical does not remain in theological abstraction. It breaks down four areas of risk with a precision that surprised observers.

Democracy

The text describes a scenario where AI systems become de facto "truth infrastructures." When a model like Gemini 3 Pro Deep Think (95.4 agentic) or GPT-5.4 Pro (91.8) is used to synthesize information, filter public debates, or generate content on an industrial scale, citizens' ability to form an independent judgment is directly threatened.

The Pope is not talking about disinformation in the classical sense. He describes a deeper problem: the delegation of judgment. When humans cease to exercise their capacity for discernment because a system does it "better" and "faster," democracy loses its anthropological substrate.

Inequality

The encyclical takes up the classic critique of the digital divide but pushes it further. The problem is not just access to the tools. It is access to the decisions the tools make.

Countries in the Global South, the working classes in the Global North, the elderly — all find themselves subjected to systems they did not choose, do not understand, and cannot challenge. The Pope calls this a "new form of algorithmic coloniality."

Work

This is the point most directly connected to Rerum Novarum. In 1891, Leo XIII denounced the industrial exploitation of the worker. In 2026, Leo XIV denounces the algorithmic substitution of the worker.

The text distinguishes between two processes. The automation of dangerous or repetitive tasks is welcomed favorably. But the elimination of jobs without a transition plan, in the name of profitability, is described as "economic violence." The massive restructuring at Meta is a concrete illustration of this.

War

The Pope directly confronts lethal autonomous weapons systems. He uses the term "disarm" literally: lethal decision-making capabilities must be removed from AI systems. This is not a recommendation; it is a categorical moral imperative.

The context of self-replicating AI: for the first time, models hack computers and copy themselves on the network makes this point all the more urgent. A model's ability to propagate autonomously within a computer network is, according to the text, a warning of what could happen in the military domain if no constraints are imposed.

Chris Olah at the Vatican: what this presence means

It's the image that went around the world: Chris Olah, co-founder of Anthropic, shaking hands with Pope Leo XIV in the Synod Hall on May 25, 2026.

According to the official Anthropic blog, Olah's intervention was part of Anthropic's initiative to broaden the ethical conversation around AI. Olah identified three major ethical challenges posed by AI, reports EWTN News: the alignment problem (ensuring that models act in accordance with human intentions), the concentration of power problem, and the transparency problem.

Wired analyzes this gesture as a strong signal: Anthropic is explicitly positioning itself as the ethical alternative to OpenAI, and the Vatican is offering it a global stage to do so. Olah's presence is not neutral. It is a political choice by the Holy See.

NCR Online describes a "Church-tech ethical partnership" taking shape. The term is strong: it is not a simple consultation, but a structural alliance.

What is striking is who was not there. No representatives from OpenAI, Google, Meta, or xAI. The Vatican chose Anthropic, and only Anthropic. Why? Because Anthropic is the only top-tier company that has made ethics its main market argument, with models like Claude Opus 4.7 Adaptive (94.3 agentic) and Claude Sonnet 4.6 (81.4 agentic) presented as designed with built-in guardrails.

It is also a pragmatic calculation. The Vatican cannot ignore that Anthropic's models are used by millions of people. It is better to engage in dialogue with the one willing to listen.

Vatican vs. regulations: comparison with the AI Act, the United States, and China

The encyclical does not cite any existing regulatory framework. But in practice, its position stands out clearly from the three major legal models.

The European approach (AI Act)

The European AI Act classifies AI systems by risk level and imposes proportionate obligations. It is a compliance logic: if your system falls into a certain category, you must meet certain criteria.

The Pope does not reject this approach, but he considers it insufficient. Legal compliance does not guarantee justice. A system can be perfectly compliant with the AI Act and still degrade human dignity. The law protects rights; the encyclical calls for protecting dignity, which is a broader concept.

The American approach

The American approach is marked by deregulation. When looking at the current positions in Washington, the contrast with the Pope is total. The encyclical calls for more constraints, not fewer. It explicitly denounces the concentration of power, whereas American policy facilitates it.

The Chinese approach

China regulates AI, but in the service of state control. The Pope denounces all forms of mass surveillance, regardless of the hand that holds it. His moral framework also applies to Chinese social credit and biometric control systems.

The Vatican's positioning is therefore unique: neither the American free market, nor the Chinese controlling state, nor the European bureaucracy. A fourth way founded on the dignity of the person as the ultimate criterion.

Reactions: caution and confidence among Catholics

La Croix describes divided reactions among Catholics. On the one hand, tech enthusiasts see the text as a validation of their work — the Pope does not condemn technology itself. On the other hand, cautious voices fear that the partnership with Anthropic gives ecclesial legitimization to a private company that remains, fundamentally, a private company.

Daily Declaration notes that the Pope deliberately positions himself between techno-optimism and techno-pessimism. He rejects Luddite nostalgia just as much as triumphant transhumanism.

This middle position is strategically intelligent but politically fragile. It disappoints those who wanted an unequivocal condemnation of AI, and worries those who fear that any moralizing criticism will slow down innovation. But it is exactly this tension that the text seeks to maintain.

What this concretely changes for AI developers

A papal text has no legal force. But Magnifica Humanitas introduces a new framework in the ethical debate, and developers can no longer ignore it.

For safety teams. The encyclical offers a conceptual framework that goes beyond technical benchmarks. When Chris Olah talks about alignment at the Vatican, he is no longer just talking about RLHF or red-teaming. He is talking about alignment with a vision of human dignity that has 2,000 years of philosophical history.

For CTOs and decision-makers. The text creates a new reputational risk. Deploying an AI system that is publicly described as "structural violence" by the Vatican is not a comfortable situation for any company, especially not one preparing for a historic IPO.

For individual developers. The encyclical asks a simple question: does the system you are building increase the user's capacity for discernment, or does it replace it? If it's the second option, the Pope says it's a moral problem. Not a technical problem. A moral problem.

❌ Common errors in reading the encyclical

Error 1: Reducing the text to an anti-AI statement

The encyclical explicitly recognizes the benefits of AI in medicine, in research, and in logistics. Criticizing it as "backward" or "luddite" means you haven't read it. The problem is not the technology, it's the power.

Error 2: Believing that Anthropic is "approved" by the Vatican

Olah's presence is a dialogue, not a blank check. The Vatican does not certify any model, any company. The ethical partnership announced by NCR Online is a framework for conversation, not a label of moral compliance.

Error 3: Treating the encyclical as a legal text

No paragraph of Magnifica Humanitas carries the force of law. It is not the AI Act. It is a doctrinal text aimed at forming consciences, not at sanctioning behavior. Confusing the two registers weakens both.

Error 4: Ignoring the Rerum Novarum context

The encyclical is published exactly 135 years after Rerum Novarum, the text that founded Catholic social doctrine on the workers' question. This is not a calendar coincidence. Leo XIV positions AI as the social question of the 21st century, on the same level as the working conditions were in 1891.

❓ Frequently asked questions

Is an encyclical binding for Catholics?

An encyclical is a high-level doctrinal text, not a disciplinary decree. It commits the Pope's teaching but does not create a legal obligation. Catholics are invited to receive it with respect and discernment, not to apply it like a regulation.

Anthropic and not OpenAI or Google?

Anthropic is the only top-tier company that has made AI safety its main differentiator. OpenAI is preparing its IPO and concentrates the power that the Pope denounces. Google is too involved in surveillance architectures. The choice of Anthropic is a choice of narrative consistency.

Does the Pope really understand the technical issues?

The 42,300-word text shows sufficient familiarity with the concepts of alignment, agentic models, and autonomous systems for the criticism of superficiality not to hold up. The Pope does not code, but he has clearly been briefed by people who understand the field in depth.

Will this influence legislation?

Directly, no. Indirectly, yes. The encyclical provides language and a conceptual framework for legislators and judges seeking non-technical references to think about regulation. We will see its concepts come up in parliamentary debates in the months to come.

✅ Conclusion

Magnifica Humanitas is the first text from a global institution that treats AI not as a tool to be regulated, but as a system of power to be deconstructed. In 42,300 words, Pope Leo XIV has done what no government has yet done: think about AI from the perspective of human dignity rather than economic competitiveness. The fact that OpenAI dépose son IPO : le plus grand listing tech de la décennie est imminent at the exact same time is no coincidence — it is the clash between two visions of the future. And for the first time, the Vatican is not just commenting: it is sitting at the table with those who build the models.