Leo XIV's Magnifica Humanitas Sets Out AI Accountability Framework

When the Vatican presents a major document, it rarely invites an AI researcher to the stage. That changed on May 25, when Pope Leo XIV presented *Magnifica Humanitas* at the Holy See with Chris Olah of Anthropic among the speakers. The 42,300-word encyclical is framed as safeguarding the human person in the age of artificial intelligence, and the presence of an interpretability researcher at its launch is a small signal worth noting: this document engaged seriously with how AI actually works before making claims about what it cannot do.

The headline argument is organized around two biblical images, the Tower of Babel and the rebuilding of Jerusalem's walls, but strip the theology and what remains is a concrete policy framework. Section 98 states that current AI systems are "more 'cultivated' than 'built,' for developers do not directly design every detail," language that maps closely onto how interpretability researchers describe emergent behavior in large neural networks. The document draws a clear line on AI consciousness: "So-called artificial intelligences do not undergo experiences, do not possess a body, do not feel joy or pain." This is not a hedge; it is a position.

The governance prescriptions are the most practically interesting part. Section 105 argues that "responsibility must be clearly defined at every stage: from those who design and develop these systems," a chain-of-accountability framing that echoes provisions in recent AI regulation. Section 108 addresses power concentration directly: "AI tends to amplify the power of those who already possess economic resources, expertise and access," and argues that "data is the product of many contributors and should not be treated as something to be sold off." Section 101 notes that "current AI systems require enormous amounts of energy and water, significantly influencing carbon emissions."

What the document does not do is name specific companies or propose enforcement mechanisms. The warnings about private technology companies wielding power surpassing many governments are pointed but general. The honest caveat is that moral frameworks from religious institutions have historically moved policy on long timescales, not quarterly ones. What the reporting does not give you is any detail on how much direct input AI researchers had in shaping the technical sections; the language about AI being "cultivated" rather than "built" suggests significant engagement, but the drafting process is not disclosed.

The encyclical roots its argument in a 135-year tradition of Catholic Social Teaching running from Leo XIII's 1891 *Rerum Novarum*. Leo XIV presented the encyclical personally at the Vatican, unusual for a papal document, which signals it is meant as a direct intervention rather than a delegated statement. The companies, legislators, and institutions the document addresses most pointedly are exactly those with least incentive to treat a papal encyclical as binding, which is either the document's fatal limitation or the reason it needed to exist.