Popes do not usually trend on Hacker News. Magnifica Humanitas cracked 1,400 points and 800 comments within a day of publication, which says something about how thoroughly the questions it raises have leaked out of philosophy departments and into the standup meetings of people who build software. Leo XIV chose his regnal name in deliberate echo of Leo XIII, whose 1891 Rerum Novarum is the foundational text of modern Catholic social teaching on industrial labor. The new encyclical is positioned as its sequel for the algorithmic economy, and it is the first papal document of its kind to engage directly with machine learning systems, recommendation engines, and the labor practices around data annotation.
I am not Catholic. I build Discord bots, write small systems tools, and ship features that depend on large language models. The document is still worth a careful read by anyone in that position, because it is one of the few sustained attempts by a major non-technical institution to articulate what it would mean for these systems to be built well rather than merely built fast.
What the document actually says
The encyclical is divided into four parts. The first reviews the anthropological argument the Church has built up since Gaudium et Spes in 1965, that human dignity is intrinsic and prior to any utility calculation. The second part applies this to machine systems, arguing that any system that mediates between persons inherits a responsibility for their dignity, whether or not its designers intended that mediation.
The third part is where developers should pay attention. Leo XIV singles out four practices by name: opaque content ranking algorithms, behavioral advertising targeting minors, automated hiring and firing decisions made without meaningful human review, and the conditions under which data labelers in the Global South are employed by AI training pipelines. He cites reporting on Kenyan workers paid under two dollars an hour to filter toxic content for ChatGPT training, and he names this as a contemporary form of what Leo XIII called the exploitation of labor without proportional reward.
The fourth part is the prescriptive section, and it is more restrained than I expected. It does not call for moratoria, does not endorse specific regulations, and explicitly rejects the framing that AI is intrinsically opposed to human flourishing. The argument is that tools are subordinate to ends, and the ends should be debated openly by the communities affected rather than settled inside product organizations.
The labor argument is the sharpest part
If you read only one section, read paragraphs 47 through 63 on what the document calls “the new invisible workforce.” Leo XIV draws a direct line from the piecework looms Leo XIII condemned to the Mechanical Turk marketplace and to the contract data-labeling firms that supply training data to most of the large model labs. The argument is structural: the further the worker is from the entity that benefits from their work, the easier it becomes to suppress wages, deny benefits, and discard the relationship at will.
This is not a new observation. The labor economist Mary Gray has been making it in academic terms since at least 2019. What is new is a globally distributed institution with roughly 1.4 billion members putting it into a foundational document. Compare this to the response from the EU AI Act, which gestures at supply chain transparency without enforcing it, or to voluntary frameworks like the Partnership on AI’s responsible data enrichment guidelines, which large labs sign but do not contractually bind their subcontractors to.
The encyclical does not solve this problem. It does insist that solving it is constitutive of building these systems well, and that handwaving about it disqualifies a project from claiming social legitimacy.
What it is careful not to say
The document avoids several positions that critics expected. It does not endorse a personhood claim for AI systems, and explicitly rejects what it calls the “category confusion” of treating language models as moral patients. It does not call for slowing AI development on existential-risk grounds; the framing throughout is about present harms to present people, not about future superintelligence. It does not name companies. It does not endorse any specific regulatory framework, though it cites approvingly the Council of Europe AI Convention as an example of multilateral effort.
This restraint is doing real work. The audience is not regulators, it is the half-billion or so people who work in or adjacent to the technology industry and who have some affiliation with the Catholic tradition. The document is asking them to bring their professional judgment into alignment with a moral tradition, rather than treating the two as separate magisteria.
What a developer can do with this
The practical question is whether the encyclical changes anything you would do on Monday morning. For most engineers it will not, in the sense that you are unlikely to refactor a sprint around it. The places it should bite are the decisions that get bundled into “that’s just how the platform works,” the ones that surface when a product manager asks for a dark pattern in the onboarding flow, or when the data team proposes a labeling contract with a vendor whose payment practices are not auditable.
For those of us building smaller things, the relevant section is paragraph 71, where Leo XIV writes that scale is not a defense. A small Discord bot that nudges users toward compulsive engagement participates in the same logic as a recommendation engine that does the same at a billion-user scale. The moral question does not require global reach to apply. That is uncomfortable, and I think it is correct.
The HN comment thread is worth reading alongside the document, because it surfaces the same arguments that surface inside engineering organizations whenever this topic comes up: that ethics frameworks are unenforceable, that the alternative to building these systems is letting worse actors build them, that abstract dignity claims do not survive contact with quarterly metrics. The encyclical does not refute these arguments. It situates them in a longer conversation about how communities have negotiated the introduction of powerful technologies before, and how those negotiations have usually involved institutions outside the industry itself.
Whether you accept the theological frame or not, that historical pattern is real. Industrial labor reform did not come from inside the factories. It came from organized labor, churches, journalists, and eventually regulators, working in uneven combination. There is no reason to expect the algorithmic economy to be different. Leo XIV’s document is one institution declaring it intends to be part of that combination. The interesting question is who else will be.