The Pope's AI Encyclical Is Really About Who Controls the Future

The first thing worth knowing about Pope Leo XIV's new encyclical Magnifica Humanitas is that it isn't really an AI document. It uses AI as a lens to diagnose something older and harder to fix: the steady capture of transformative technology by a very small group of people who were never elected, rarely scrutinized, and have no particular obligation to the rest of us.

If you're an engineer, student, or builder in Sri Lanka, that framing should feel immediately familiar. You're building with tools whose rules are written by people who didn't consult you.

---

🔍 What the Encyclical Actually Says

The document runs to roughly 200 pages and was presented alongside Chris Olah of Anthropic — a detail worth noting. The Vatican invited someone from inside one of the leading AI labs. Whether that signals genuine dialogue is a fair question.

The central argument, reported by TechCrunch on 25 May 2026, is that "when power is concentrated in the hands of a few, it tends to become opaque and evade public oversight." The encyclical cites concrete recent examples: Elon Musk's acquisition of Twitter, tech elites funding super PACs to block regulation, and the reported influence of venture capitalist David Sacks on the Trump administration's AI executive orders.

Key takeaway: The pope isn't warning us that AI will become sentient and destroy humanity. He's warning that a handful of unelected people are currently deciding how the world's most powerful technology gets governed — and that this is a structural problem, not a technical one.

The comparison it draws is deliberate. Magnifica Humanitas references Rerum Novarum, Pope Leo XIII's 1891 response to the Industrial Revolution. Industrial power was also concentrated, also opaque, and also governed primarily by whoever could build the machines. The current moment rhymes. What changed is the speed and the scale.

---

🌐 The AI Arms Race Critique

One of the document's sharper positions is a call to end the AI arms race — the competitive pressure that pushes labs, states, and companies to ship fast and defer governance questions. This is what most public AI debates sidestep. The problem isn't any individual model or capability. It's the competitive dynamics that make safety and accountability feel like friction rather than requirements.

Here's roughly where different actors sit on the power-to-accountability axis right now:

Actor	Influence over AI direction	Public accountability
Major AI labs (OpenAI, Anthropic, Google DeepMind)	High	Low — private or barely-public
Nation-state regulators (EU, US, UK)	Medium	Medium
International bodies (UN, ITU)	Low	High, but largely toothless
Developers and builders globally	Very low	N/A

That last row is where most of us sit. We consume the APIs, build on the platforms, fine-tune on the models — and have essentially no voice in how the underlying infrastructure gets governed. The encyclical's core objection is to exactly this arrangement.

---

🧠 The Cognitive Freedom Problem

Notre Dame law professor Paolo Carozza is quoted in the TechCrunch piece as saying AI-driven misinformation has "corroded our capacity to recognize what's true," posing "fundamental challenges to cognitive freedom."

This is the piece that lands hardest for working engineers. It's not only that bad actors spread misinformation. It's that systems we build and use are optimizing for engagement at the expense of epistemic quality — and most of us doing the building don't control that optimization target. The product managers do. The business model does.

If the tools you build end up being used to corrode people's ability to think clearly, who bears responsibility? The encyclical's answer: everyone in the chain, but especially those who set the direction.

This is less a theological argument than a design ethics argument. And it applies directly to anyone in Sri Lanka building AI-adjacent products — recommendation feeds, content surfaces, chatbots, automated summaries — where the engagement-versus-accuracy tradeoff is entirely real and rarely examined.

---

💡 What "Oversight With Community Participation" Actually Looks Like

The encyclical calls for "clear criteria and effective oversight" with community participation in AI governance. Easier said than done — but a few practical readings for builders:

• Open-weight models (Llama, Mistral, Gemma) reduce the information asymmetry between independent developers and the concentrated few who control frontier AI. When weights are public, anyone can audit, fine-tune, and evaluate them. This isn't just a cost saving; it's a governance choice.
• Regulatory frameworks like the EU AI Act, whatever their imperfections, represent an attempt to move governance decisions out of boardrooms and into accountable institutions. Following their development matters even if your product targets Sri Lanka rather than Europe.
• Community-level accountability — researchers publishing model evaluations, journalists investigating training data practices, civil society groups pushing for transparency — is exactly the "community participation" the encyclical endorses. You can contribute to this without waiting for a policy seat at any table.

None of this involves endorsing a specific regulation or theology. It's a framework point: whoever holds power must be visible and accountable. We've largely failed to apply that to technology companies, and the encyclical is pointing at the gap.

---

🛠️ What This Means for You

If you're developing AI-adjacent products in Sri Lanka, the practical takeaways aren't about religion. They're about how you build:

1. Document your optimization targets. If your product uses AI to surface content or generate text, write down what's being maximized, who controls it, and what users can do when the output is wrong. This takes an hour and surfaces assumptions you've been carrying silently.

2. Prefer open infrastructure where your budget allows. Running open-weight models locally means your project doesn't depend on governance decisions made in boardrooms you have no access to. It's also free at inference, which matters on a learning budget.

3. Track governance developments, not just capability benchmarks. The EU AI Act, UNESCO's AI ethics framework, and local regulatory signals all affect what you can ship and how. These move slowly but their impact lands hard when they do.

4. Treat neutrality as a design decision, not a default. Every AI tool has values embedded in its design — usually the values of whoever funded the training run. Using the tools with that awareness doesn't mean refusing them; it means being honest with your users about what they're interacting with.

The pope's document runs 200 pages and is addressed to 1.4 billion people. But its core concern scales down cleanly: power without accountability tends to become a problem, whatever the technology. Whether you're building a three-person startup product in Colombo or reading this as a student thinking about what to build next, that's a principle worth starting with.

---

Source: TechCrunch — "The pope's AI encyclical isn't really about AI", published 25 May 2026.