War is easy. Civilization is hard.

There is a brutal asymmetry at the heart of politics and war.

Creating value is hard. Destroying it is easy. Take value from others, and you do not have to create it yourself.

A city takes decades to build. A missile can ruin part of it in seconds. A functioning institution depends on trust, discipline, competence, and time, while a few corrupt people, a coup, or a war can hollow it out very quickly. A business, a culture, even a generation of social stability can be painstakingly assembled and casually burned. People who do this do not necessarily have to be evil in some cartoonish sense. Often they do it because they benefit, because they manage to capture some of the created value for themselves.

That does not make destruction productive in any deep sense. It is parasitic. It makes the world poorer overall. Yet this is exactly why the problem keeps returning: an action can damage the whole and still reward the actor. If someone gains power, territory, leverage, prestige, or a larger share of what remains, then net destruction may still look like success from inside their incentives.

That is one of the oldest civilizational problems. The people who create value usually need patience, cooperation, long time horizons, and some degree of peace. The people who seize value need much less. Force, timing, leverage, and willingness are often enough.¹

The private logic of a public loss

From the outside, war often looks absurd. In the aggregate, it usually is. It burns lives, infrastructure, capital, trust, and future possibilities. But if you zoom in on the level of the individual actor, the faction, the regime, or the empire, the logic becomes less mysterious.

A lot of conflict is rational in relative terms while remaining irrational in civilizational terms.

That is the same pattern you see in simpler games. In a prisoner's dilemma, the individually tempting move can produce the collectively worse outcome. In a common goods game, everyone benefits if the shared resource is sustained, yet each participant is tempted to defect, extract a bit more, and hope somebody else carries the burden. If that second term is unfamiliar, the closely related idea most people know is the tragedy of the commons. Politics works like that far more often than people want to admit. The shared resource is not just a pasture or a public fund. It is social trust, legal stability, productive capacity, civic peace, and sometimes an entire country.

This is why societies so easily drift toward predation. Producing value asks for competence and restraint. Capturing value can be done by interruption. Tax it, extort it, confiscate it, monopolize it, inflate it away, sanction it, sabotage it, or bomb it. Even when the total pie shrinks, the actor who gains relative advantage may still count it as a win.

That is also why we should be wary of social systems, not only of obviously violent individuals. Systems can reward extraction just as reliably as people can choose it. A bureaucracy can become parasitic. A ruling class can prefer a poorer society it controls over a richer one it would have to share. A military apparatus can become self-justifying. Institutions that are meant to protect can turn into institutions that feed on the thing they administer.²

Still, that is not an argument against institutions as such. It is an argument for better ones.

Civilization as protection for creators

If building is harder than taking, then civilization is, among other things, the attempt to make creation safer than predation.

That is what property rights, courts, checks on arbitrary power, stable money, civil liberties, and boring administrative competence are really for. They are not decorative luxuries for peacetime. They are mechanisms that protect builders from takers, so people can invest in the future without assuming that tomorrow somebody stronger will simply seize the result.

A decent society is not one in which nobody has power. That has never existed. A decent society is one in which power is constrained enough that the engineer, teacher, founder, researcher, farmer, artist, and worker can create value with some confidence that it will not immediately be consumed by organized extraction.

This is why strong governance and healthy suspicion of power have to coexist. We need institutions precisely because unstructured competition tends to reward defectors and raiders. At the same time, those institutions have to be shaped so they defend the productive base instead of preying on it. Civilization is not the abolition of force. It is the taming of force, the channeling of it, and the constant refusal to let it become the highest principle.

AI is a tool with two edges

This is where AI enters the picture.

The first thing worth saying is that AI is not automatically a war technology. It is a general amplifier. It can help people reason, coordinate, summarize, detect patterns, model systems, and navigate complexity. In principle, those capacities can serve war, but they can also serve medicine, education, science, law, infrastructure, and public administration.

That is exactly what makes the technology so consequential.

If you want a better civilization, AI could help with many of the things that modern states and institutions often do badly. It could make legal systems more legible, help citizens understand regulation, assist judges and lawyers with massive bodies of precedent, surface corruption patterns, detect procurement anomalies, improve public planning, support dispute resolution, and reduce bureaucratic friction that currently wastes everyone's time. It could help smaller institutions gain competence that previously only giant bureaucracies or elite firms could afford.

In that sense, AI could become civilizational infrastructure. Used well, it might lower the cost of fairness, competence, and coordination.

But tools do not arrive in a vacuum. They arrive in power structures, and those power structures do not think only about justice. They think about control.

How AI can make war easier

Once you look at AI through that lens, the danger becomes obvious.

If destruction and extraction were already often cheaper than creation, AI can widen that gap by compressing the cost of surveillance, classification, targeting, propaganda, and command support. What once required large teams of analysts and administrators can increasingly be assisted by systems that sort signals, flag anomalies, rank entities, fuse information, and produce action-ready summaries at scale.

None of this requires a science fiction scenario where killer robots suddenly wake up. The more realistic danger is more banal and therefore more serious: existing institutions get better tools for seeing, sorting, influencing, and acting.

A military can use AI to process intelligence faster. A security state can use it to correlate behavior across huge data sets. A propaganda apparatus can use it to generate narratives, test them, personalize them, and flood the information environment. A targeting pipeline can move from sensing to recommendation to action with less delay, less uncertainty, and less room for hesitation.

We already have concrete examples of how dangerous that becomes when bad data gets folded into operational software. In its reporting on the 2026 bombing of the Shajareh Tayyebeh primary school in Minab, the Guardian argued that the real issue was not a rogue chatbot at all, but Project Maven, by then developed by Palantir into targeting infrastructure, combined with an outdated Defense Intelligence Agency database that still treated the site as a military facility. CNN and NPR separately reported the stale intelligence record and satellite evidence showing the school had long since been converted from its earlier adjacent military use.⁷

That last part matters. Much of human restraint lives in friction. Delay gives time for doubt, review, argument, and conscience. Systems built to reduce friction do not only remove inefficiency. They also remove moral buffers.

This is why so many arguments about keeping a human in the loop feel weak. If the machine sets the rhythm, frames the options, and floods the operator with ranked outputs, then the human may still be present formally while becoming hollow in practice. Decision support turns into decision pressure. Oversight survives on paper, but real judgment starts slipping away.³

The fight over lines

That is why lines matter, and why they are so hard to hold.

Science fiction has actually been useful here. Asimov's three laws of robotics were never a real engineering solution, but they captured something important: once we build systems with increasing agency, we start needing explicit principles for where force, obedience, and safety should stop.⁴ Older AI discourse was full of warnings like do not connect the thing to the internet, do not let it self-propagate, do not let automation outrun supervision. With large language models, we blew past a lot of that caution almost instantly.

Now the real question is no longer whether we should draw lines. It is whether we still can.

The recent Anthropic conflict is revealing for that reason. The interesting part is not some simplistic story of an AI company refusing all defense work. By Anthropic's own account, it was willing to support lawful national security use, but wanted to keep two red lines: mass domestic surveillance and fully autonomous weapons.⁵ For that, it was reportedly labeled a supply chain risk.⁶

That should make people stop and think.

If even narrow limits of that kind are treated as unacceptable friction, then the pressure is clearly toward tighter integration of AI into coercive systems. The real contest is not AI in war, yes or no. That question is already over. The real contest is whether democracies and institutions retain any meaningful boundary between analysis and action, surveillance and control, assistance and autonomy.

A civilization worthy of intelligence

It would be a mistake, though, to end the story there and sink into pure pessimism.

The same capacities that make AI attractive to militaries and surveillance systems could also be used to strengthen the conditions under which people are able to build. Better forecasting could help avoid conflict. Better institutional memory could reduce bureaucratic stupidity. Better legal tools could help ordinary people defend themselves. Better civic interfaces could make governance more understandable and participatory. Better audit systems could expose predation earlier, whether it comes from states, corporations, or criminal networks.

A civilization worthy of intelligence would use intelligence to protect creators, not only to optimize controllers. It would ask how technology can widen access to competence, distribute understanding, reduce corruption, and make it easier to cooperate across difference. It would not confuse the ability to monitor everything with the wisdom to govern well.

That also points to a broader cultural task. If society feels like it is slipping, the answer cannot only be better weapons, tighter control, or louder ideology. We need common understanding, shared purpose, local trust, and communities that remember how to create together. We need institutions people can believe in, but also living networks of family, friendship, neighborhood, craft, and mutual aid. We need more builders, and we need cultures that honor building.

War will probably never disappear completely. Neither will greed, fear, ego, or the temptation to seize rather than create. But civilization has always been the wager that these impulses do not have to rule everything. We can make cooperation more attractive, predation more costly, and human dignity harder to crush.

That is still the task now.

In the end, unity does not mean pretending conflict is unreal. Hope does not mean refusing to see danger. It means deciding, again and again, that the world we want will not be built by the people who are best at breaking things. It will be built by people who can create value, defend one another, and hold the line long enough for something worth living in to grow.

Footnotes