When AI Becomes a Workplace Mandate:

Pope Leo XIV’s Magnifica Humanitas should not be read as another cautious Vatican reflection on technology. It is more ambitious than that. It is an attempt to relocate the debate about artificial intelligence from the shallow language of tools, disruption, and productivity into the older and more dangerous language of dignity, freedom, truth, work, and conscience. The encyclical’s central claim is not that AI is evil. It is that AI forces us to ask what kind of human beings our institutions are forming, what kinds of power they are legitimating, and what kind of obedience they are beginning to demand.

Within days of the encyclical’s publication, people online started using it as a possible basis for a religious exemption from employer-required AI workflows. The reels online used captions that made the connection clear through hashtags such as #aiworkflow, #religiousexemption, and #corporatesocialresponsibility. The visible comments showed that viewers immediately understood the practical implications: some joked about becoming Catholic, others imagined texting HR, and others insisted that at-will employment would make the argument futile.

That reaction may look unserious at first. It is not. It is the first public test of whether Magnifica Humanitas will remain a theological document about technology or become a living source of institutional friction. The real question is whether any employee can now refuse any AI tool. That would be too crude. The more profound question is whether a worker can say, with moral seriousness, "I am being asked not merely to use software, but to participate in a sociotechnical system that violates my conscience, damages human agency, hides responsibility, or treats persons as optimization objects."

The distinction matters. It moves us from the superficial question, “Is AI allowed by Catholic teaching?” to the deeper institutional question, “Under what conditions does AI serve or degrade the human person?” It moves us from “Can an employer require AI tools?” to “When does a workplace mandate become participation in a morally contested system?” It moves us from treating refusal as personal discomfort to asking whether refusal can express a coherent religious or moral claim about dignity, freedom, and responsibility. It also moves HR away from a narrow compliance posture and toward the harder question of whether existing religious-accommodation frameworks are ready for objections to AI workflows.

The answer is not obvious, and it should not be made obvious too quickly. If the post-pandemic workplace taught employers anything, it is that conscience claims do not remain inside seminar rooms. They appear in the form of policies, job descriptions, Slack messages, and litigation risk. AI is now entering the same territory.

The encyclical’s argument: AI is not neutral because institutions are not neutral

The first mistake would be to read Magnifica Humanitas as a Vatican warning against machines. Pope Leo XIV says something subtler. Technology is not an enemy of humanity, and it is not inherently evil. Yet it is always socially charged, because it takes on the characteristics of those who design, finance, regulate, deploy, and use it. This is why the encyclical does not ask whether AI is impressive. It asks whether AI is being ordered toward the common good or toward concentration, domination, and dependency.

That distinction places Pope Leo XIV in a long philosophical tradition. In “The Question Concerning Technology,” Martin Heidegger argued that we are most endangered by technology precisely when we treat it as merely neutral. His point was not that machines have intentions. Modern technology discloses the world in a particular way, as material to be ordered, optimized, stored, and used.

“Everywhere we remain unfree and chained to technology, whether we passionately affirm or deny it. But we are delivered over to it in the worst possible way when we regard it as something neutral.”

Pope Leo's encyclical translates a similar insight into Catholic social language. The human person is not merely an input. Work is not merely throughput. Truth is not whatever circulates efficiently. Responsibility is not discharged by placing a human somewhere near a dashboard. If AI is deployed inside institutions that already reward speed, scale, opacity, surveillance, and cost reduction, it will not magically humanize those institutions. It will amplify their moral grammar.

Ellul’s The Technological Society did not define technology narrowly as machines. He used the term "technique" to describe the totality of rationalized methods ordered toward efficiency. The danger of technique is that it quietly makes efficiency the governing criterion before anyone has adequately examined the ends. Pope Leo’s “technocratic paradigm” is a close theological cousin of Ellul’s concern. Once efficiency becomes sovereign, the question “Can we automate this process?” begins to replace “Should this human relation be mediated, measured, or delegated?”

Heidegger, Ellul, Illich, and Pope Leo do not make the same argument, but they converge on one decisive point. Heidegger helps us see that technology changes how reality appears to us, which is why AI can make workers, customers, and knowledge appear as data resources. Ellul helps us see that efficiency can become a ruling social logic, which is why AI mandates often smuggle moral judgments into productivity metrics. Ivan Illich offers the constructive alternative: tools should enlarge human competence, initiative, and control, not convert the person into an accessory of a system. Pope Leo then gives this philosophical concern a theological and social form: technology must be governed by dignity, truth, work, freedom, and the common good.

Illich is especially important because he prevents the argument from becoming merely negative. In Tools for Conviviality, he argued that tools become humane when they enlarge personal competence, control, and initiative, but become destructive when they turn people into accessories of bureaucracies or machines. That is almost exactly the axis on which the encyclical’s AI argument turns. AI is acceptable where it remains a tool for human flourishing. It becomes morally suspect when it reorganizes human beings around machine-readable priorities.

The theological core: human dignity is not productivity by another name

The encyclical’s theological foundation is not a decorative preface. It is the entire argument. The human person, Pope Leo insists, is created in the image of God and therefore possesses a dignity that is not acquired, earned, optimized, ranked, or justified by output. Human dignity is ontological, not performative. It belongs to the person before performance and after failure.

This matters because AI systems usually enter institutions through measurable promises. They promise faster drafting, cheaper support, improved prediction, better risk detection, more targeted communication, and more efficient decision-making. None of these goods is trivial. Yet the moral danger is that institutions begin to understand the worker, the patient, the student, the applicant, or the citizen according to the categories that AI systems can process.

In Catholic terms, this is an anthropological reduction. In secular terms, it is a category error. A human being can be evaluated in some contexts, but cannot be exhausted by evaluative systems. The Vatican’s earlier note Antiqua et Nova makes the same distinction with particular clarity. AI may perform tasks associated with intelligence, but human intelligence belongs to the whole person, including embodiment, moral judgment, affectivity, relationality, and openness to truth.

Antiqua et Nova distinguishes task performance from human thought: AI systems can perform sophisticated tasks, but they do not possess intelligence in the full human sense because human intelligence belongs to the person as an embodied, relational, moral, and spiritual being.

That distinction prevents two opposite errors. The first is technophobia, which imagines every use of AI as a betrayal of humanity. The second is technocratic naivete, which imagines that if AI can imitate an activity, the human meaning of that activity has been captured. The encyclical rejects both. It does not deny the usefulness of AI. It denies that usefulness can settle the question.

What Magnifica Humanitas actually demands

The document’s moral framework is broader than AI ethics checklists. It retrieves Catholic social doctrine as an integrated way of reading institutional life. The key principles are the common good, the universal destination of goods, subsidiarity, solidarity, and social justice. Each principle tests AI at a different level.

The common good asks whether an AI system promotes shared human flourishing or private extraction. In workplace terms, productivity gains cannot be the only measure of success. The universal destination of goods asks who controls the data, infrastructure, models, and computational power that increasingly mediate social life. If a small number of firms, employers, or states concentrate AI capability, they will deepen dependency and exclusion. Subsidiarity asks whether agency is preserved at the lowest competent level. Workers should not be reduced to passive executors of automated decisions. Solidarity asks who bears the hidden cost of the system, including annotators, displaced workers, monitored employees, affected communities, and those whose lives are rendered legible only as data. Social justice asks whether vulnerable persons are protected before harm occurs, which means AI governance must shape design and deployment rather than merely manage reputational damage after the fact.

The encyclical therefore demands more than polite transparency. It calls for accountability across the lifecycle of AI, independent oversight, user education, contestability, legal frameworks, protections for minors, data protection, supply-chain responsibility, worker participation, retraining, and corporate metrics that include the quality and dignity of work.

That last phrase is crucial. If a company adopts AI and then measures success only by speed, headcount reduction, or output volume, it has already failed the encyclical’s test. The Catholic tradition does not treat work as merely a cost center. John Paul II’s Laborem Exercens states that work bears “the mark of a person operating within a community of persons" and that the Church must defend the dignity and rights of workers amid technological and economic change.

The point is not nostalgia for pre-digital labor. The point is that the worker is a subject, not an instrument. AI becomes morally dangerous when it reverses that order, when the human worker becomes the monitoring appendage, prompt supplier, liability absorber, or emotional cover for a system whose purposes were set elsewhere.

Why “human in the loop” may not be enough

Many organizations answer AI concerns with a reassuring phrase: there will be a human in the loop. The phrase is better than full automation, but it can also be morally evasive. The question is not whether a human is present in the loop. The question is whether the human has meaningful authority, sufficient understanding, practical time, institutional permission, and moral responsibility proportionate to the decision.

Madeleine Clare Elish’s concept of the moral crumple zone helps us here. In aviation and automation contexts, Elish describes cases in which human operators absorb blame for failures in systems over which they have limited control. The human becomes the moral buffer for automation. The machine remains impressive, the institution remains protected, and the nearest human takes responsibility for a failure that the system has structured.

Cognitive science gives the same warning from another angle. Automation bias is the tendency to over-rely on automated outputs, even when contrary evidence is available. A 2024 report from Georgetown’s Center for Security and Emerging Technology argues that human oversight can be weakened by interfaces, institutional pressures, training environments, and the social authority granted to automated systems. Legal scholars Johann Laux and Hannah Ruschemeier have likewise argued that automation bias is not simply an individual psychological flaw. It is technical, social, legal, and normative.

The danger is that the language of oversight can preserve the appearance of responsibility while hollowing out its conditions. Human oversight can become human rubber-stamping under time pressure. AI assistance can lead to deskilling and dependency. Faster decisions can mean the loss of deliberation and contestability. Accountability can become blame-shifted to the nearest human operator. Efficiency can silently redefine an institution's ends by privileging whatever can be optimized.

This is why employer-mandated AI use is not morally equivalent to adopting a new email client. Some tools alter the medium of work. Others alter the worker’s agency inside the work. If AI drafts, ranks, recommends, filters, scores, nudges, or escalates decisions, the employee may become entangled in judgments with a partially hidden moral structure.

Could religious accommodation law take this seriously?

In the United States, the legal structure is not imaginary. Title VII prohibits religious discrimination in employment and requires reasonable accommodation of sincerely held religious beliefs, practices, and observances, unless such accommodation would impose undue hardship. In Groff v. DeJoy, the Supreme Court clarified that undue hardship means a burden that is substantial in the overall context of the employer’s business, not merely a trivial or de minimis cost.

The Court held that an employer must show that granting the accommodation would result in “substantial increased costs in relation to the conduct of its particular business.”

That does not mean AI refusal automatically wins. It means the inquiry cannot be dismissed as absurd at the threshold. HR Dive’s 2026 reporting already notes that attorneys are seeing AI-related religious objections concerning environmental impact, energy infrastructure, loss of autonomy in human decision-making, worker displacement, and unethical uses such as warfare. The same article notes that if an employee does not truly need the AI system to perform essential job functions, a simple accommodation might be exemption from organizational AI-use requirements. If AI is essential to the job, or if avoiding it would make performance impossible, the analysis becomes harder.

The analogy to biometric hand-scanner litigation is instructive. In EEOC v. Consol Energy, an employee objected on religious grounds to using a hand scanner. The point is not that AI is equivalent to the “mark of the beast.” It is that an unusual technology-related religious objection may still receive legal attention if the belief is sincere and a reasonable alternative exists.

In practical terms, the strongest claims will likely be concrete rather than absolute. A writer who asks not to use generative AI for drafts, while still meeting the same quality and productivity standards manually, presents a more serious accommodation case than someone who objects to any workplace process touched by AI. A designer who objects to uploading original work into AI systems for training or generation raises a significant conscience question, especially if intellectual labor is being absorbed into opaque systems. A customer service employee who refuses an AI assistant that determines answers, tone, and escalation presents a fact-specific question: Does the role still permit meaningful human judgment without the tool? A worker who objects to AI used for surveillance, ranking, or discipline may raise a particularly strong moral concern when human dignity, autonomy, and due process are implicated. By contrast, a blanket objection to all contact with products, services, or processes touched by AI becomes much harder if avoidance would restructure the entire job or the employer’s operating model.

This is where the encyclical sharpens the debate. A weak objection says: “I do not like AI.” A stronger objection says, "This required workflow makes me cooperate in a system that treats human judgment as replaceable, hides responsibility, exploits labor, increases surveillance, or causes unjust harm.” The second claim may still fail legally in some circumstances. But it is not trivial.

Specific projections: what this will likely change

The next two years will probably produce a new category of HR policy: AI conscience accommodations. These will not be as common as disability accommodations or vaccine-era religious requests, but they will not be rare in knowledge-work sectors. The first wave will likely involve writers, educators, designers, lawyers, healthcare workers, customer support agents, software developers, and HR professionals whose work requires them to either generate AI content, evaluate people using AI systems, or upload human work to AI platforms.

Employers will respond unevenly. Sophisticated organizations will distinguish between optional AI assistance, role-essential AI systems, and prohibited AI uses. Less careful organizations will announce broad “AI-first” mandates, then discover that a universal productivity slogan becomes fragile when confronted by religious accommodation, intellectual property, labor, privacy, and discrimination questions.

A second projection follows from the moral crumple-zone problem. As AI systems become more agentic, employers will increasingly demand that employees remain accountable for outputs they cannot fully inspect. This will intensify conscience claims. Workers will not merely object to using AI. They will object to being made the responsible human face of a system whose recommendation logic, training data, environmental cost, labor chain, or downstream harm they cannot evaluate.

A third projection concerns religious pluralism. Catholic workers may be the first visible group to invoke Magnifica Humanitas, but they will not be the only ones. Environmental religious traditions, Orthodox Jewish and Muslim concerns about human agency and accountability, Protestant worries about deception and idolatry, Buddhist concerns about attention and suffering, and secular moral-conscience objections may all converge around different aspects of AI. The workplace will not be debating “religion versus technology.” It will be debating whether institutions can compel participation in systems whose moral meaning is contested.

These projections suggest several likely institutional consequences. AI conscience accommodations will emerge because AI mandates are spreading faster than ethical governance, requiring employers to create intake processes, role analysis, and alternatives. Creative and professional workers will raise many of the first serious claims because their work is directly transformed by generative systems, forcing policies to distinguish assistance, replacement, and appropriation. Accountability objections will intensify because agentic AI widens the gap between output and human control, placing “human in the loop” language under scrutiny. Religious arguments will increasingly merge with labor and privacy arguments because AI systems affect dignity, autonomy, surveillance, and displacement together. The first court cases will likely be messy, sincere, fact-specific, and technologically complex, with early precedents turning less on broad AI theology than on job necessity, available alternatives, and the concrete burden placed on employers.

None of this requires an anti-AI posture. It requires refusing the childish idea that AI adoption is a one-way civilizational maturity test. A worker can be technically competent and still morally opposed to particular AI workflows. A Catholic can use AI in one context and refuse it in another. A company can adopt AI and still owe employees serious explanations about what systems do, what data they consume, what judgments they shape, and who remains responsible.

The question the encyclical leaves us with

The genius of Magnifica Humanitas is that it does not let us hide inside either panic or enthusiasm. It asks whether AI helps us remain human. That question is harder than asking whether AI helps us move faster. It forces institutions to account for the kind of human agency they are preserving or destroying.

The Instagram reel matters because it gives us the first public application of that question in ordinary workplace language. If an employer says AI is now part of the job, can a worker respond that this is not merely operational but spiritual, not merely technical but moral, not merely a matter of preference but a matter of conscience?

The law will not answer that question abstractly. It will ask about sincerity, essential functions, reasonable alternatives, undue hardship, costs, operational context, and consistency. Theology will not answer it mechanically either. It will ask whether the worker is resisting real moral participation in a dehumanizing system or simply baptizing discomfort with change.

But the question itself is now unavoidable.

Can religious freedom be used to compel an employer not to force us to use AI?

Not always. Not never. Not as a slogan. Not as a trick.

But perhaps it is a serious claim when AI use becomes participation in a system that violates conscience, obscures responsibility, diminishes human judgment, or treats the person as a standing reserve for a technocratic order.

That is the question Magnifica Humanitas has placed on the table. It is now a Vatican, HR, legal, and human rights question. The most honest thing to say is that we do not yet know how far it will go. The most dishonest thing would be to pretend it has not already begun.

References

[1] Pope Leo XIV, *Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence*, Vatican, May 15, 2026.

[2] Isabella Piro, “Pope Leo’s ‘Magnifica humanitas’: AI must serve humanity not concentrate power,” Vatican News, May 25, 2026.

[3] Instagram reel by `skyfisherforskyfish`, Reel DY0fSX0PBug, accessed May 28, 2026. Analysis is based on visible caption, hashtags, comments, and readable subtitle fragments available through the public page.

[4] Martin Heidegger, “The Question Concerning Technology,” in *The Question Concerning Technology and Other Essays*, translated by William Lovitt, 1977.

[5] Jacques Ellul, *The Technological Society*, translated by John Wilkinson, New York: Vintage Books, 1964.

[6] Ivan Illich, *Tools for Conviviality*, 1973.

[7] Dicastery for the Doctrine of the Faith and Dicastery for Culture and Education, *Antiqua et Nova: Note on the Relationship Between Artificial Intelligence and Human Intelligence*, Vatican, January 28, 2025.

[8] John Paul II, *Laborem Exercens*, Vatican, September 14, 1981.

[9] Madeleine Clare Elish, “Moral Crumple Zones: Cautionary Tales in Human-Robot Interaction,” *Engaging Science, Technology, and Society* 5, 2019.

[10] Lauren Kahn, Emelia S. Probasco, and Ronnie Kinoshita, “AI Safety and Automation Bias: The Downside of Human-in-the-Loop,” Center for Security and Emerging Technology, November 2024.

[11] Johann Laux and Hannah Ruschemeier, “Automation Bias in the AI Act: On the Legal Implications of Attempting to Debias Human Oversight of AI,” *European Journal of Risk Regulation* 16, no. 4, 2025.

[12] U.S. Equal Employment Opportunity Commission, “Religious Discrimination,” accessed May 28, 2026.

[13] Supreme Court of the United States, *Groff v. DeJoy*, 600 U.S. 447, 2023.

[14] Ryan Golden, “AI mandates may stir up religious objections. HR should prepare now,” HR Dive, May 4, 2026.