Disarming AI

Paragraph 110 in full

The full text, from the Vatican's English translation:

"Finally, I would like to employ the expression 'to disarm,' which is close to my heart. Disarming AI means freeing it from the mentality of 'armed' competition, which today is not limited simply to the military context, but is also an economic and cognitive phenomenon. This entails a race for ever more powerful algorithms and larger datasets, driven by the desire to secure geopolitical or commercial dominance. To disarm means discrediting the assumption that technical power automatically confers the right to govern. To disarm does not mean rejecting technology, but preventing it from dominating humanity. It means freeing technology from monopolistic control and opening it to discussion and debate, therefore making it human-friendly and restoring it to the plurality of human cultures and ways of life. Our task today is not only ethical or technical. It is ecological in the deepest sense, for it concerns a new dimension of our common home. AI is already an environment in which we are immersed, as well as a force with which we must engage. For this reason, merely regulating it is insufficient; it must be disarmed, welcoming and accessible."

Magnifica Humanitas, n. 110

Read slowly, the paragraph is doing five things in sequence. It introduces the term. It expands the meaning of "armed competition" beyond the military to include the economic and cognitive. It rejects the equation of technical power with governing authority. It distinguishes disarming from rejecting. And it concludes with the strongest claim in the passage: regulation alone is insufficient, because the AI environment now constitutes "a new dimension of our common home" and must be addressed at that environmental level.

The five moves are not equally well-known. The "armed competition extends beyond military" claim is the one most quoted in early reception. The "regulation is insufficient" claim is the one most likely to land hardest in policy circles, because it explicitly rejects the dominant frame of contemporary AI governance.

Why the phrase matters

The standard AI policy vocabulary has three main terms: AI safety, AI alignment, and AI ethics. Each does specific work, and each has limits the introduction of disarming AI was designed to address.

AI safety is the engineering frame. The problem is treated as a technical risk profile (capability, robustness, deception, misuse) and the solution as better engineering practice (red-teaming, evaluations, deployment controls, capability evaluations). The frame is useful and necessary. It is also limited. AI safety as a discipline operates inside the assumption that the AI development cycle, with its current actors and incentives, is the cycle within which safety has to be achieved. Safety work happens inside frontier labs, regulated by frontier lab leadership, evaluated by frontier-lab-adjacent institutions. The frame cannot easily ask whether the development cycle itself ought to look different.

AI alignment is the calibration frame. The problem is treated as ensuring that AI systems pursue goals aligned with human values or human intentions, and the solution as technical work on training procedures, reward modeling, and constitutional methods. Like AI safety, alignment is useful and necessary. Like AI safety, it operates inside the assumption that the question of what should be built is logically prior to the question of how to align what is built — and that the first question is settled, leaving only the second to address.

AI ethics is the principles frame. The problem is treated as the absence of clear normative guidance for AI development and deployment, and the solution as the articulation of principles (fairness, accountability, transparency, beneficence) to be applied. The frame is in some sense the broadest of the three, but it has been criticized from inside the AI ethics community itself for being abstract, easily co-opted, and disconnected from the actual decisions that shape AI's effects in the world.

Disarming AI does work none of these terms does. It names the development cycle itself, with its competitive race dynamics, as part of the problem to be addressed. Paragraph 110's argument is that "the race for ever more powerful algorithms and larger datasets, driven by the desire to secure geopolitical or commercial dominance" is not a neutral condition inside which AI safety, alignment, and ethics work happens. It is itself a condition that has to change. A technically safe, well-aligned, ethically principled AI system built inside a race economy will still produce outcomes the race economy produces: concentration of power, exclusion of communities that cannot keep up, treatment of human cultures as input data, the displacement of judgment by speed.

This is why the phrase will outlive the news cycle. The AI policy world has been searching for a vocabulary that names the structural problem, not just the technical one. "Race to the bottom" gets close but is too generic. "AI capture" gets close but is too specific to regulatory capture. "Disarming AI" names the right object — the competitive escalation dynamic itself — with the right register: serious, demanding, and connected to a long tradition of thinking about how to prevent technologies from dominating the societies that develop them.

The three meanings of disarmament

Paragraph 110 names three dimensions of "armed competition" that disarming AI is meant to address. Each has its own register and its own implications.

The military meaning

AI as a weapon. The encyclical's Chapter 5 develops this dimension at length (paragraphs 197 to 200) and reaches what may be the document's sharpest single claim: "No algorithm can make war morally acceptable" (paragraph 198). Lethal autonomous weapons systems, AI-augmented targeting, AI-driven decision support in conflict situations, and the broader integration of AI into the military-industrial complex are the immediate referents.

The military meaning is the most legible to most readers because it sits inside an existing tradition of disarmament discourse. The Church has worked on nuclear disarmament for sixty years; the application to AI is in some sense a natural extension. Pope Leo XIV made the parallel explicit at the launch event: "The Church has long been working for nuclear disarmament, aware that every great technical power can affect people's lives and so must be accompanied by adequate moral discernment and public control. Nuclear disarmament remains a service to peace and the dignity of the human family. In a similar sense, artificial intelligence now demands to be disarmed, freed from logics that turn it into an instrument of domination, exclusion and death."

The military meaning is the place where disarming AI is least controversial. Most major religious traditions, many secular ethics frameworks, and a growing portion of the AI safety community converge on the claim that lethal decisions should not be delegated to autonomous systems. The encyclical's distinctive contribution at this dimension is less the content of the claim and more the magisterial weight behind it.

The economic meaning

The commercial race among AI labs for more powerful systems, more compute, more data, more capable models, and greater market position. The encyclical's argument is that this race functions as a form of armed competition in its own right, with its own escalation dynamics, its own concentration of power, and its own displacement of common goods.

The economic meaning is the one that has the sharpest implications for the AI industry. The current frontier-lab development pattern — private companies racing each other to release more capable systems faster, funded by investors with explicit return expectations, operating under competitive pressure that incentivizes corner-cutting on safety, alignment, and deployment caution — is precisely the pattern the disarmament frame names as part of the problem.

Anthropic co-founder Christopher Olah's remarks at the Vatican launch event are notable in this connection. Olah, speaking alongside Pope Leo XIV, made the strongest public statement by a major AI lab co-founder on the limits of industry self-governance: "Every frontier AI lab operates inside a set of incentives and constraints that can sometimes conflict with doing the right thing." Olah named four sources of pressure (commercial viability, the drive to remain on the research frontier, geopolitical pressure, and "human pride and ambition") and concluded that outside scrutiny is essential. Read against the disarmament frame, Olah's remarks are an acknowledgment from inside the industry that the economic dimension of armed competition is real and that the industry cannot resolve it alone.

The cognitive meaning

The geopolitical and intellectual competition for control of AI as a foundational capability. Paragraph 110 names this dimension as the cognitive expression of the same underlying competitive structure: states and blocs racing to secure dominance over the infrastructure (compute, data, models, talent) that will shape the next several decades of economic and political power.

The cognitive meaning is the most expansive and the least developed in the document. The encyclical's broader argument in Chapter 3 (especially paragraphs 95 and 108) names the asymmetry between actors with access to compute and data and actors without that access as a new form of structural inequality. The disarmament frame extends this analysis to the level of the global system: a world in which AI development is structured as a competition between great powers will produce different outcomes than a world in which AI development is structured as a shared project of the human family.

The cognitive dimension is where the disarmament frame is most ambitious and most contested. It implies that the current geopolitical structure of AI development — with U.S., Chinese, and (to a lesser extent) European competition as the dominant structural feature — is itself a problem to be addressed, not just a fact to be worked within. This is a claim with no obvious operationalization. The encyclical does not propose specific multilateral arrangements. It points to the direction.

The three dimensions are explicitly connected. The encyclical does not treat them as independent problems. The military disarmament call cannot be separated from the economic disarmament call, because the commercial race produces the capabilities that get military-adapted. The economic call cannot be separated from the cognitive call, because the geopolitical competition is what funds and pressures the commercial race. Disarmament, in the encyclical's framing, is integrated or it is incomplete.

Theological roots: the Church's disarmament tradition

The phrase "disarming AI" does not arrive from nowhere. The Church has a developed tradition of thinking about disarmament that runs from John XXIII's Pacem in Terris (1963) through the U.S. bishops' 1983 pastoral letter The Challenge of Peace through the Holy See's leadership on the 2017 Treaty on the Prohibition of Nuclear Weapons. Pope Francis pushed the tradition further, calling the possession of nuclear weapons "immoral" rather than merely conditionally acceptable. Pope Leo XIV's "disarming AI" extends the same tradition into a new technological domain.

The structure of the tradition matters. Catholic disarmament thinking is not pacifism in the strict sense. It does not claim that all force is impermissible. It claims that certain categories of technology, by virtue of their scale, their indiscriminate effects, or their resistance to moral judgment, generate conditions in which legitimate use becomes impossible to specify. Nuclear weapons crossed that threshold for the Church not because all weapons are forbidden but because nuclear weapons cannot be used in ways that satisfy the moral requirements (proportionality, discrimination, last resort) the just-war tradition demands.

"Disarming AI" applies the same structure of analysis. The claim is not that all AI is impermissible. The claim is that AI as currently developed and deployed, with its concentration of power, its opacity to public oversight, its embedded competitive dynamics, and its accelerating capabilities, generates conditions in which the standard ethical resources (alignment, safety, ethics principles, post-hoc regulation) are insufficient. The disarmament call is the recognition that the threshold has been crossed and that the response has to be commensurate.

This places the AI question inside a much larger conversation than the contemporary AI policy debate has been holding it inside. The contemporary debate has been structured around governance, safety, and economic policy. The disarmament frame places AI inside the same category as nuclear weapons, biological weapons, and other categories of technology where the magnitude of potential harm and the difficulty of controlling deployment generate a structural problem the standard tools cannot resolve.

Whether this analogical move is correct is a question for the long reception. The early indication is that it will be contested. Some readers will argue that the analogy overstates the case — that AI, however powerful, is not categorically equivalent to nuclear weapons, and that the disarmament frame is rhetorically powerful but analytically loose. Other readers will argue that the analogy understates the case — that AI's pervasiveness and its effects on the conditions of human cognition put it in a category nuclear weapons do not occupy. The encyclical does not resolve the debate. It opens it.

What disarming AI asks of governments

The encyclical does not propose specific policy mechanisms. It is not that kind of document. But the disarmament frame, combined with the encyclical's broader analysis in Chapters 2 and 3, points to a set of policy directions that governments would have to take seriously to engage the call.

Robust legal frameworks over ethics in the abstract. Paragraph 106 is direct: "It is not enough to invoke ethics in the abstract; robust legal frameworks, independent oversight, informed users and a political system that does not abdicate its responsibility are required." The disarmament call is incompatible with self-regulation as the primary governance mode. It requires binding law and binding oversight.

Treatment of data as a common good. Paragraph 108 develops the argument that data is "the product of many contributors and should not be treated as something to be sold off or entrusted to a select few." The disarmament frame extends this to the broader infrastructure of AI: compute, models, training data, and the algorithmic systems built on them are not naturally private goods. Their concentration in private hands is a policy choice that the disarmament call asks governments to reconsider.

Restoration of public authority over decisions currently delegated to private platforms. Paragraph 95 names the asymmetry: control over platforms, infrastructure, data, and computing power does not rest with states but with major economic and technological actors that "effectively set the conditions for access, determine the rules of visibility and shape the very possibilities for participation." The disarmament call asks governments to recover authority over decisions that are properly public, including decisions about how AI systems shape employment, credit, public services, and democratic discourse.

International cooperation that curbs the technological arms race. Paragraph 200 names the requirement directly in the context of weapons systems: "It is imperative to establish a shared framework — also at the international level — in order to curb the technological arms race and ensure robust protection for civilians." The disarmament frame extends this to the economic and cognitive dimensions: international frameworks for the development of AI, not just for its weaponized uses, are part of what the disarmament call asks.

None of these directions is operationally specified. The encyclical is a magisterial document, not a regulatory text. The work of translating these directions into specific policy mechanisms is the work of governments, multilateral bodies, civil society, and the policy community. The disarmament frame is meant to shape that work, not to substitute for it.

What disarming AI asks of developers

Paragraph 111 follows immediately on the introduction of the disarmament call, and it addresses AI developers directly. The address is unusual in a social encyclical. Most papal documents speak in general terms to all the faithful and all people of good will. Paragraph 111 speaks specifically to one professional class:

"I wish to address a special appeal to those who develop artificial intelligence. In one sense, technological innovation can represent human participation in the divine act of creation. Developers, therefore, bear a particular ethical and spiritual responsibility, for every design choice reflects a vision of humanity. Just as the creator of an artistic or literary work must consider the values it conveys, so developers are called to embed values in their projects with due seriousness: with transparency, responsibility toward affected communities and careful attention to ensuring that what is being cultivated is a genuine good."

Magnifica Humanitas, n. 111

Three things in this passage are worth attention. First, the framing is theological elevation, not condemnation. Developers are addressed as participants in a creative act with the dignity that implies, not as ethical risks to be managed. Second, the responsibility named is "ethical and spiritual," not just professional. The encyclical is asking developers to read their work through a frame that most contemporary AI development does not provide for them. Third, the practical content is concrete: transparency, responsibility to affected communities, attention to whether what is being built is a genuine good.

Read against the disarmament frame, paragraph 111 is doing specific work. The disarmament call could be heard as an indictment of the AI industry as such. Paragraph 111 forecloses that reading. The encyclical is not asking developers to stop developing. It is asking developers to develop differently — with the kind of moral attention that, in the encyclical's account, the current competitive structure makes difficult to sustain. The address is an invitation: read yourself into this document as a participant in the work of disarming, not as the target of it.

For developers reading the encyclical seriously, the question is what disarming AI looks like at the level of daily professional practice. The encyclical does not provide a checklist. It points to dispositions: transparency about design choices and their effects, responsibility that extends to the communities downstream of the system, attention to the actual goods the work serves rather than the proxy metrics that dominate the industry's evaluation systems. These are not new ideas in AI ethics, but the encyclical's framing gives them magisterial weight and connects them to a much larger tradition than the contemporary AI ethics conversation has typically reached for.

What disarming AI is not

Several misreadings are predictable. Naming them prevents them.

Disarming AI is not a call to reject AI. Paragraph 110 says explicitly: "To disarm does not mean rejecting technology, but preventing it from dominating humanity." The encyclical's broader argument in Chapter 3 is that technology is not in itself evil and can serve genuine human goods. Readers who interpret the disarmament call as a Luddite gesture have misread the document.

Disarming AI is not the same as AI safety or AI alignment. The three vocabularies are not opposed, but they are not equivalent either. AI safety and alignment work can proceed without disarming AI; disarming AI cannot be achieved through safety and alignment work alone, because the disarmament frame names problems (concentration of power, competitive escalation dynamics, displacement of public authority) that safety and alignment frames are not designed to address. Practitioners can pursue both, and many will. They should not be confused.

Disarming AI is not a specific policy proposal. The encyclical does not propose particular regulatory mechanisms, particular treaty arrangements, or particular institutional structures. It names a direction. The translation of that direction into specific policy is work that has to happen at the level of governments, multilateral bodies, civil society, and the policy community. Readers who expected an action plan have misread the document's genre.

Disarming AI is not exclusive to Catholics. The encyclical is addressed to "all men and women of good will," in the standard magisterial form. The disarmament frame draws on Catholic resources, but it is offered as a contribution to a conversation the whole human family is already having. Non-Catholic readers can engage the frame on its merits without accepting its theological foundations. The frame's analytical content — that the AI development cycle itself, not only its outputs, is part of what has to change — stands or falls on its own terms.

Disarming AI is not primarily about the military meaning. The military meaning is the most legible because of the existing disarmament tradition. But paragraph 110 is explicit that armed competition "is not limited simply to the military context, but is also an economic and cognitive phenomenon." Readers who reduce the disarmament call to lethal autonomous weapons have taken the most familiar dimension for the whole concept. The full frame is broader, and its broader application is where most of the work has to happen.

Adopting the frame: practical guidance

For journalists, homilists, teachers, researchers, and practitioners considering how to use the phrase in their own work:

Cite paragraph 110 specifically. The phrase has its definition there. References to "disarming AI" without the paragraph citation will be harder to evaluate and easier to dismiss. Standard form: Pope Leo XIV, Magnifica Humanitas, n. 110.

Use all three meanings. The phrase loses most of its distinctive content if it is reduced to the military dimension. The economic and cognitive dimensions are what make the frame more than an application of existing disarmament thinking. Treatments that retain all three meanings will be more accurate to the document and more useful in their own analysis.

Pair the phrase with paragraph 106 and paragraph 200. Paragraph 106 grounds the disarmament call in concrete governance demands ("robust legal frameworks, independent oversight"). Paragraph 200 specifies the international dimension. The phrase, taken alone, can read as gestural. Paired with these adjacent passages, the operational implications become clear.

Hold the distinction with AI safety and AI alignment. The phrase does specific work that those terms do not. Conflating them will reduce its analytical content. Practitioners working in AI safety or alignment can engage the disarmament frame without abandoning their own work; the frames are complementary, not substitutable.

Connect to the Church's longer disarmament tradition where useful. The theological background is real. Treatments that connect "disarming AI" to Pacem in Terris, to the Holy See's role in the Treaty on the Prohibition of Nuclear Weapons, and to Pope Francis's intensification of the Catholic position on nuclear weapons will be stronger than treatments that present the phrase as an isolated coinage.

Be honest about what is contested. The frame is not universally accepted, even inside Catholic ethics. Some readers will argue the analogy with nuclear weapons is overdrawn. Treatments that engage these challenges seriously will be more credible than treatments that present the frame as settled doctrine.

What to watch over the next year

The trajectory of "disarming AI" as a phrase will be visible in three places.

Catholic intellectual reception. Whether the phrase gets adopted by Catholic universities, episcopal conferences, theological journals, and Catholic AI ethics centers will determine whether it acquires the institutional weight that turns a coinage into a tradition. The early signs from the launch event and the Catholic intellectual press are positive. The deeper test will be whether the phrase appears in working documents from places like Notre Dame, Georgetown, Santa Clara, Boston College, the Catholic University of America, and the Pontifical Academy for Life over the next twelve months.

Secular AI policy uptake. Whether the phrase reaches AI policy conversations outside the Catholic context is the harder test. The early indication is mixed. Mainstream press has quoted the phrase but not always centered it. Major AI policy institutions (Brookings, CSIS, the various AI safety institutes) have not yet engaged it formally. Whether the phrase becomes a serious term of art in secular AI policy or remains a primarily Catholic vocabulary will be visible by the end of 2026.

Industry response. Whether AI developers, especially at the frontier labs, engage the phrase will be the most consequential test. Christopher Olah's remarks at the Vatican launch suggest that at least one frontier lab co-founder is willing to engage the disarmament frame substantively. Whether others follow, whether the phrase becomes part of how AI developers describe their own work, and whether it shapes the actual practice of AI development at the frontier will determine whether "disarming AI" ends up doing the work the encyclical is asking it to do.

None of these tests will be settled quickly. The phrase has been in public for less than a day. Its trajectory will be visible by the end of the year and consequential by the end of the decade. What is clear now is that the phrase has been introduced with the magisterial signal that marks the addition of a serious new term to the Catholic conversation, and that the reception so far has been substantial enough to suggest the phrase has the durability to become what the encyclical is asking it to become.