Pope Leo’s AI Warning Is Really About Platform Power
Pope Leo XIV used his first encyclical, Magnifica Humanitas, to put artificial intelligence at the center of a political and moral argument about power. NPR, citing Religion News Service and AP photos from the Vatican, reported that the document warns AI could widen inequality, weaken democracy and undermine human dignity if development is shaped mostly by military competition, economic pressure and Big Tech incentives. Leo called for AI to be “disarmed,” a phrase he defined not as rejecting technology but as preventing it from dominating humanity. The encyclical argues that AI amplifies the power of those who already control capital, expertise and data, allowing small influential groups to shape information, consumption patterns, democratic processes and economic outcomes. The timing is fresh because AI regulation has usually been framed as a technical safety debate or a race with China. Leo’s intervention reframes it as a governance question: who gets to set the defaults for a technology that is rapidly entering schools, workplaces, politics, war and everyday life.
Pope Leo’s AI encyclical is interesting not because the Vatican suddenly became a technology regulator. It is interesting because it names the thing most AI debates try to avoid: power.
The normal conversation around artificial intelligence is trapped between two sales pitches. One says AI is an unstoppable productivity miracle, so governments should get out of the way. The other says AI is an existential danger, so experts and regulators should be trusted to manage it from the top down. Both frames leave ordinary citizens in the same place: outside the room where the defaults are set.
Leo’s language cuts through that. He is not saying the technology should be banned. He is saying that a technology this powerful cannot be treated as neutral just because it arrives inside an app, a cloud contract or a workplace dashboard. AI does not simply “happen.” It is trained on data, financed by capital, deployed through platforms, purchased by governments and optimized for incentives. That means the politics are built in before the public ever sees the product.
That is the real platform-power story. The largest AI companies are not just selling software. They are building the operating layer for work, education, media, surveillance, advertising, medicine and eventually parts of state administration. If a handful of firms control the models, the data pipelines, the compute infrastructure and the distribution channels, then they are not merely participating in the market. They are shaping the choice architecture of society.
This is where the usual left-right fight becomes too small. The question is not whether AI is “good” or “bad.” The question is whether the benefits and risks are being socialized honestly. If AI lowers costs, who captures the gains? If AI replaces tasks, who absorbs the transition? If AI filters information, who audits the filter? If AI tools become embedded in policing, defense, hiring, credit, insurance or health triage, who is accountable when the system makes a consequential mistake?
Leo’s warning about inequality matters because the technology rewards incumbency. Data favors incumbents. Compute favors incumbents. Distribution favors incumbents. Regulatory complexity can also favor incumbents, because the biggest firms can afford compliance teams while smaller challengers cannot. That creates a nasty political loop: the companies most capable of shaping the rules are also the companies most likely to benefit from rules written around their own architecture.
The phrase “disarm AI” will sound dramatic to some readers, but the useful part is not the slogan. The useful part is the reminder that competition can become militarized even outside the battlefield. Companies race for dominance. Governments race for strategic advantage. Investors race for monopoly returns. Workers, teachers, parents and local communities are told to adapt after the fact.
A serious AI politics would not start with panic. It would start with institutional humility. It would ask where human review is mandatory, where automated decisions should be banned, where open competition needs protection, where procurement needs transparency, and where citizens deserve to know when a machine is shaping the answer they receive.
That is not anti-technology. It is pro-accountability. The AI boom may produce real abundance. It may also produce a narrower society where a few firms own the rails and everyone else rents access to the future. The difference will not be determined by model benchmarks alone. It will be determined by who has leverage when the rules are written.