AI in the Age of Tragic Realism.

Three Documents⌗

In May 2026, three documents appeared within days of each other. Together, they reveal a contradiction at the heart of the AI moment — one that none of them can resolve. And if we don’t name it honestly, we will sleepwalk into a future that nobody chose.

Pope Leo XIV published Magnifica Humanitas, his first encyclical, arguing that humanity faces a choice between building Babel and rebuilding Jerusalem. Anthropic co-founder Chris Olah stood at the Vatican and asked for “moral voices the incentives cannot bend.” And Anthropic itself published a geopolitical strategy paper arguing that American democracy must maintain AI dominance over China through export controls, anti-distillation enforcement, and global distribution of the American AI stack.

Read together, these documents don’t form a synthesis. They form a fracture. And we need to stare into it.

The Encyclical’s Novel Move⌗

Magnifica Humanitas does something genuinely new. It takes the principle of the universal destination of goods, the old idea that the earth’s resources belong to all humanity, and applies it to patents, algorithms, data, and digital platforms. It takes subsidiarity, the principle that decisions should be made as close as possible to the people affected, and extends its reach. The threat to subsidiarity now comes not only from the overreaching state, but from the private technology company that defines the conditions of everyday life for billions while answering to almost no one.

These are not rhetorical gestures. They are substantive intellectual moves that reorient a centuries-old moral tradition toward the actual structure of power in 2026. We should take them seriously, because almost nobody else is saying this with institutional weight behind it.

The Olah Paradox⌗

Eleven days before Chris Olah asked the Pope for critics that “the incentives cannot bend,” his own company published a paper whose strategic logic is structured by commercial and geopolitical incentives. The paper frames AI competition as a two-scenario contest: either America wins decisively, or authoritarianism prevails. It advocates tighter export controls, crackdowns on model distillation, and making American AI “the backbone of the global economy.”

At the Vatican, Olah asked: how can we ensure the gains of AI are shared globally? The policy paper answers: by ensuring they are not shared with more than one billion Chinese people.

I don’t think this is simple hypocrisy. It might be something worse — a genuine tension that the people inside the system can feel but cannot resolve. Olah’s speech reads like someone who sees the moral stakes clearly and also knows the machine he’s part of is moving in a direction those stakes can’t fully govern. That honesty matters. But honesty without structural change is just eloquence.

Containment’s Leaky Premise⌗

Here’s what troubles me about the policy paper: its own evidence undermines its theory of victory. Chips are smuggled. Data centers in Southeast Asia provide remote access to export-controlled compute. Distillation attacks continue despite enforcement. The paper calls for closing these loopholes, but the history it documents is a history of loopholes opening faster than they close.

If containment cannot be made airtight — and nothing in the paper suggests it can — then we are asking the wrong question. It is not “how do we maintain dominance?” It is “what institutions and norms do we want in place when our lead inevitably narrows?” That is exactly the question the encyclical is built to answer. And it is exactly the question the policy paper refuses to ask.

We are spending our strategic window on building higher walls when we should be spending it on building durable institutions. Walls crack. Institutions, if built well, outlast the power dynamics that created them.

The Domestic Narrative Lock⌗

But the encyclical has its own blind spot, and it is the one that keeps me up at night.

The adversarial dynamic between the US and China may not be solvable by moral appeal or strategic dominance, because it may not be primarily about the relationship between the US and China at all. The CCP’s posture toward the West serves a domestic function: it sustains the narrative of civilizational revival against hostile encirclement that justifies single-party rule. Remove the external threat and you weaken one of the internal justifications for the regime.

This means Western actions are always interpreted through a domestic lens. The West could adopt every principle in the encyclical — share AI universally, dismantle export controls, build perfectly inclusive multilateral institutions — and the CCP would still frame it as a trap, because the framing serves an internal need that has nothing to do with Western behavior.

The encyclical assumes adversarial dynamics result from moral failure. The policy paper assumes they can be managed through dominance. Neither framework fully accounts for one mechanism that keeps reproducing the conflict: a domestic political structure that requires an external enemy to function.

Aristotle’s Warning⌗

This is where I find myself reaching past all three documents toward something older.

In the Politics, Aristotle observed that all regime types decay — not through accident, but through their own internal logic. Monarchy slides toward tyranny. Aristocracy toward oligarchy. Democracy toward mob rule. The decay is driven not by bad actors but by the structural tendency of every regime to hollow out its own justifying principle.

Look at what’s happening. Democracy organized around permanent technological dominance over an external enemy is already deteriorating. When “democratic values” become branding for a system where a handful of private actors hold more power than most governments, we are watching polity slide toward oligarchy in real time. The regime label stays. The substance rots.

And authoritarianism armed with AI makes internal correction harder. The surveillance and control capabilities that frontier AI enables loosen one of the historical constraints on authoritarian rule: its dependence on human enforcers whose loyalty could waver.

AI does not create the concentration-of-power problem. It accelerates it in both systems simultaneously. The question is not whether AI serves democracy or authoritarianism. It is whether any regime’s balancing mechanisms can keep up with the speed of concentration that AI enables.

I fear they cannot.

If I Were President Xi⌗

Here is the part that should make us all uncomfortable.

If I sat in Beijing and read these three documents, I would not see a dilemma. I would see an opportunity. The encyclical gives me the moral language to challenge American AI dominance at every international forum — universal access, solidarity, subsidiarity, the common good. The policy paper tells me exactly where my adversary’s theory of victory is vulnerable: a containment regime that leaks, a legitimacy claim that rings hollow in the Global South, a concentration of power in private hands that contradicts the democratic values it invokes. And the Vatican speech tells me that at least some people inside the American AI industry know the contradiction is real.

I would weaponize the encyclical’s language of universal sharing while building my own closed system. I would offer the Global South cheaper, less restricted AI infrastructure and frame American export controls as the new colonialism. I would enthusiastically support multilateral governance frameworks, the slower and more inclusive the better, buying time to close capability gaps while looking like the reasonable party. I would stop trying to win the frontier race and focus instead on deploying good-enough AI at scale across my military, industrial base, and domestic apparatus, where my advantage is speed of integration, not model intelligence.

None of this requires genius. It requires only a competent leader reading publicly available documents.

But — and this is the point I want to press hardest — this is not a story about Xi Jinping hating the West, nor about Xi’s personal psychology. I don’t think hatred enters into it. Any leader sitting in that chair — managing a single-party state whose legitimacy depends on civilizational revival against perceived external containment — would make substantially similar moves. The configuration requires it. The domestic political architecture needs the adversarial posture to function, regardless of what the adversary actually does or intends.

That is what makes this genuinely tragic rather than merely competitive. Xi does not need to hate America. He needs the narrative of American containment to hold his own system together. And America, by pursuing containment, supplies exactly the evidence that narrative requires. The policy paper advocates tightening export controls. Beijing points to tightened export controls as proof of hostile encirclement. The population coheres around the Party. The cycle tightens.

No amount of moral appeal from the Vatican breaks this loop. No amount of strategic pressure from Washington breaks it either. The adversarial dynamic is not a misunderstanding to be resolved. It is a structural feature of the current configuration of power. Both sides are locked inside it. And AI is making the lock stronger.

Tragic Realism⌗

So where does this leave us? Not where any of us would like.

The encyclical’s moral framework is correct as a standard of judgment. The policy paper’s threat assessment is largely accurate as a description of present conditions. But neither offers a solution, because the underlying problem — structural regime decay compounded by an adversary whose internal architecture requires the conflict — may not have one.

The honest position is not the encyclical’s hope for moral transformation, nor the policy paper’s confidence in strategic dominance. It is something more modest and more painful: damage limitation. Buying time. Building whatever partial institutions are possible while the window holds. Maintaining the standard of human dignity as a measure by which both Washington and Beijing are found wanting, in different ways and at different speeds.

The encyclical warns against building Babel. The policy paper describes a Babel with democratic branding. The adversary is building its own Babel while citing the encyclical’s language of universal solidarity. The question is whether anyone is willing to stop.

Leo XIV reaches for theology — for grace, for the Magnificat, for a transformation that politics alone cannot deliver. Whether that is sufficient depends on commitments that precede any policy analysis. But as a diagnosis, the encyclical may be the most durable of the three documents, because it names the one thing the others cannot say: that the problem is not which faction holds power, but what power does to every faction that holds it.

That is the older warning Aristotle helps recover. That is what the Church, at its best, has always insisted. And it is the only honest starting point for thinking about AI, geopolitics, and what it means to remain human while building machines that increasingly resemble us.

The choice between Babel and Jerusalem is being made now, with every chip we restrict, every model we deploy, every institution we build or fail to build.

I don’t have the answer. But I know the first step is to stop pretending anyone else does either.