Pope Leo XIV released an encyclical on artificial intelligence on May 25, covering the environmental toll of large models, the risks of automated decisions affecting individual lives, and the concentration of power that AI creates for those who already hold economic resources. Simon Willison, the developer and blogger, published a detailed breakdown on his site the same day, calling it some of the clearest writing he has seen on the ethics of integrating AI into modern society.
An encyclical is a formal teaching document issued by the Pope, carrying doctrinal weight for the Catholic Church’s 1.4 billion members. It is not legislation. No tech company faces a compliance deadline because of it. What it does do is introduce a moral framework into the policy conversation at a scale and with an institutional credibility that no single AI safety organization can match. Pope Leo XIII’s 1891 encyclical Rerum novarum on labor rights in the industrial age shaped Catholic social teaching for more than a century and informed secular labor law debates across multiple continents. Pope Leo XIV chose his name precisely to invoke that precedent.
The document’s substantive claims are specific. It states that current AI systems require enormous amounts of energy and water, with significant carbon emissions implications. It warns that consequential decisions about employment, credit, and access to public services risk being fully delegated to automated systems that lack compassion, mercy, and the capacity to recognize that people can change. It names the power-concentration problem directly: AI amplifies the leverage of those who already possess economic resources, expertise, and data access, allowing small but influential groups to shape information flows, influence democratic processes, and steer economic dynamics to their own advantage.
Compare this to the existing landscape of AI moral statements. The UNESCO Recommendation on the Ethics of AI, adopted in 2021, covers similar ground on human rights and environmental sustainability. The Asilomar Principles from 2017 addressed safety and beneficial outcomes from a primarily technical community. Anthropic and OpenAI each publish safety commitments and model cards that acknowledge alignment risk. The encyclical differs from these in its framing: it is not a voluntary industry commitment, and it does not originate from the institutions being evaluated. Its authority derives from a tradition of social teaching that predates Silicon Valley by several centuries.
The structural limitation is real. An encyclical binds Catholic teaching, not corporate behavior. Nothing forces a cloud provider to reduce data center water consumption, or requires a hiring platform to add human review to algorithmic rejection decisions. The document explicitly calls for data to be treated more as a public good than a private asset, a position that conflicts directly with the business models of every major AI company. No enforcement mechanism accompanies that call.
The signal worth watching over the next twelve months is institutional uptake inside Catholic-aligned organizations. Catholic universities, hospital networks, and government ministries in Catholic-majority countries now have a first-party doctrinal document to cite when drafting AI procurement policies. That is a different kind of leverage than a press release. Italy, Spain, Brazil, Mexico, Poland, and the Philippines each have significant Catholic institutional infrastructure and active debates about AI governance.
Founders building AI products for those markets should expect specific procurement language referencing the encyclical to appear in RFPs and vendor assessments within the next year. The question is not whether a given product passes a theological test. The question is whether it can demonstrate human oversight of automated decisions, transparent accountability chains, and evidence of environmental cost management. Those are now the stated criteria of a buying constituency that did not previously have a unified document to point to.
Posted by Simon Willison on simonwillison.net on 2026-05-25.
