Reader, I didn't marry her. (Apologies to Chas* for the misappropriation, but the cadence works.) On a normal day I'm an atheist; on a bad day, agnostic. My only brush with the Catholic Church was an engagement to a wonderful woman from Campania that didn't make it to the altar, and the Tuesday (maybe Monday, my memory escapes me) evening marriage preparation classes at St Peter's Italian Catholic Church in Clerkenwell that came with it, led by the quietly formidable Padre Carmelo. The last thing I wanted after a full day's work was to sit in a church hall being guided through the sacrament of matrimony by a Catholic priest; and yet, somehow, those evenings had a warmth to them that I've never quite forgotten. I went back years later to thank him, only to find he'd moved on to another diocese. I hope he's well. I respect the beliefs of most people (with the exception of Brexiteers) and fully understand why the opiate of the people brings comfort and meaning to billions. I tend, if I'm honest, to believe in little green men before a supreme celestial being.
I say all of this upfront because what follows is not a dismissal of Magnifica Humanitas.
What the document actually is
Magnifica Humanitas runs to 245 paragraphs and addresses "safeguarding the human person in the time of artificial intelligence." Leo presented it personally at the Vatican alongside Christopher Olah, co-founder of Anthropic. Here was the head of the world's largest Christian denomination sitting beside one of Silicon Valley's leading AI researchers, signaling that the Church intends to be part of this conversation as a participant with something substantive to say.
And it does have something substantive to say. The title namechecks AI but the document's real preoccupations run deeper: inequality, the erosion of democracy, the concentration of power and the normalisation of war are what Leo is actually writing about. These are not new problems — they are old ones that AI is turbocharging. This is a document rooted in Catholic social doctrine stretching back to Leo XIII's Rerum Novarum in 1891, updated for an era in which the people concentrating power aren't industrial factory owners but the architects of algorithmic infrastructure.
Power and the data irony
The encyclical's most pointed argument is structural: the governance problem with AI is inseparable from the ownership problem. Leo calls for data to be treated as a common good rather than a private commodity, owned and regulated on behalf of everyone.
There's an irony worth naming here. Every major LLM currently operating was built by treating data as exactly that, hoovering up vast quantities of it from across the internet without asking permission or sharing the proceeds. In my view, much of this was done on legally questionable grounds; the litigation is still playing out. The difference is that the Church's version of a common good implies shared benefit and democratic oversight; the industry's version involved scraping first and litigating later.
Leo also draws a direct line from the exploitation embedded in AI supply chains to historical slavery. Millions are engaged in data labeling, content moderation and model training, frequently young people and predominantly women, working under demanding conditions for minimal wages, alongside children extracting rare earth materials in dangerous conditions to produce the hardware the whole system runs on. He issues a formal apology for the Church's own delayed condemnation of slavery and its historical role in enabling it, becoming the first pope to acknowledge that his predecessors gave European sovereigns explicit authority to subjugate and enslave those they deemed outside the faith.
That's a significant statement. But it comes from an institution whose own record on power, transparency and accountability is, to put it charitably, complicated. The Vatican Library remains largely inaccessible to the public; the history of institutional self-protection over accountability is long and documented. And the Church is not operating from a position of financial modesty. The central Holy See reported €1.23 billion in income in 2024, and the Cologne archdiocese alone holds €3.35 billion in assets. When an institution of that scale condemns the concentration of wealth and power and calls for transparency, it is speaking from inside a system that has its own obligations to examine. That doesn't invalidate the argument, but it shapes how you hear it.
The optics
Which brings me back to Olah, and to the image that most people who didn't read all 245 paragraphs will have seen: the Anthropic co-founder, seated beside the Pope, at the Vatican.
I appreciate that including technology companies in the conversation is arguably more sensible than speaking about them from a distance. Olah's own remarks were candid. He acknowledged the real possibility of AI displacing human labor at very large scale and admitted that companies like his operate inside commercial and geopolitical incentives that can conflict with doing the right thing, making outside scrutiny essential. Fine.
But Anthropic is not the AI industry. I use Gemini; I use ChatGPT; I use Claude, and if I'm honest, Claude has always felt the most comfortable of the three, though I'm aware that's subjective and how many LLMs does one person actually need. The point is that OpenAI wasn't there. xAI wasn't there. Mistral wasn't there. The document being launched was a critique of concentrated technological power, and it was launched with one AI company in the room. For everyone who won't read the full text, the takeaway is simple: Anthropic, the Pope's choice. That's an extraordinary piece of implicit endorsement, intended or not.
This isn't a new problem. When Pope Francis attended the G7 in 2024 to discuss AI, I had the same unease. His perspective was legitimate and worth hearing, but the G7 nations are home to millions of Muslims, Jews, Protestants, Buddhists, Hindus and countless other belief systems, none of which had a comparable seat at the table. The Catholic Church speaks for 1.4 billion people; that leaves several billion more whose frameworks for thinking about human dignity and technological power were not in the room. Ethical AI governance requires a much broader range of voices, not a single religious institution however thoughtful, and not a single AI company however willing to sit beside a pope.
The deeper complication
Anthropic is also not simply a tech company with progressive instincts. It has built a substantial government business. Claude is deployed across US federal agencies, in classified networks and at National Laboratories, for use cases including intelligence analysis, operational planning and cyber operations. Its primary defence distribution partner has been Palantir; that relationship became politically sensitive in 2026 when Anthropic and the Pentagon clashed over whether Claude could be used without restrictions in ways Anthropic said could enable mass domestic surveillance or fully autonomous weaponry. That dispute reportedly led the government to limit Anthropic's access, with Palantir unwinding parts of its Claude-based defence software as a result.
Anthropic has drawn a stated line at domestic surveillance and fully autonomous lethal systems; Leo's encyclical draws exactly the same line, arguing that it is not permissible to entrust lethal or irreversible decisions to artificial systems and that the "just war" theory has been rendered outdated. That alignment may be genuine. But the optics of a company simultaneously embedded in classified US national security infrastructure, seated beside the Pope as he calls for AI to be disarmed, are complicated to say the least.
Meanwhile, President Trump delayed signing an executive order that would allow the government to evaluate AI models before their release, saying he didn't want anything getting in the way of America's lead. The Pope and the President are operating from entirely different premises about what AI governance is for.
The bottom line
Magnifica Humanitas is a serious document that deserves serious engagement. The arguments about power, data, labour, war and human dignity are well made. Tech journalists, devout Catholics and religious scholars will read it and, I hope, take it on board.
My concern is not with what the encyclical says. It's with what the launch communicated: a single institution, however large and well-intentioned, positioning itself as the moral anchor for a conversation that belongs to all of humanity, alongside a single AI company that is anything but a neutral actor in the landscape the document describes.
* Charlotte Brontë