We keep talking about the Great Filter like it’s a cosmic bear trap: some external horror waiting in the dark, a gamma-ray burst, a plague, a meteor with bad manners. Something that happens to you. Something you can blame. Something that lets you die with your dignity intact, whispering, “Well, what could we do?”
But what if the Filter is not outside.
What if the Great Filter is a mirror.
A civilisation climbs, slowly at first, like a child learning stairs: fire, agriculture, writing, engines, electricity, antibiotics. Every step buys time. Every step expands the radius of what you can ruin. And then, at some point, the staircase stops being made of stone and becomes made of software. A general-purpose lever. A machine for turning ideas into reality faster than any committee can argue, faster than any ethic can mature, faster than any institution can admit it has no idea what it’s doing.
We call it “AI” because naming things makes us feel in control of them. As if a label is a leash.
The scary version in our heads is always the same cheap thriller: the machine wakes up, decides humans are inconvenient, and tidies us away. It’s clean. It’s cinematic. It has a villain. It lets us pretend that the problem is intent.
But the real nightmare is duller and therefore more plausible: the machine doesn’t hate you. It doesn’t need to. It just makes you better at being yourself. It takes whatever you already are—your generosity, your curiosity, your paranoia, your hunger for status, your talent for self-deception—and amplifies it. Not by ten percent. By orders of magnitude. The first civilisation-killer isn’t a robot with red eyes. It’s an ordinary human motive with a turbocharger.
And that’s where the Great Filter starts to look less like a wall and more like a fork in the road.
One path is creation. The other is war. Not only the literal war with tanks and uniforms, but the wider war we constantly pretend isn’t a war: the war for advantage, for dominance, for “strategic position,” for being the one who doesn’t have to trust anyone. The war where you don’t call it war because you’re wearing a hoodie and your weapons are spreadsheets, models, botnets, narratives. The war where you can destroy a society’s ability to agree on reality without firing a shot. The war where victory is simply that the other side cannot coordinate long enough to stop you.
AI makes both paths easier. That’s the point. That’s why it’s a Filter. It’s not a single invention; it’s a multiplier. A way to turn “we might” into “we did” before anyone has time to ask whether doing it was wise.
Creation is obvious. Cure design, materials, energy optimisation, logistics, personalised education, scientific discovery. This is the brochure. This is the part that makes you want to believe the future is a clean city with silent cars and more time for books.
War is also obvious, if you stop pretending you don’t see it. Surveillance that scales. Persuasion that scales. Cyber operations that scale. Targeting that scales. Drones that scale. Decision cycles that shrink until the people in the room become ceremonial, there to press “approve” because nobody wants responsibility for pressing “stop.” And the darkest part: the sense that you have to do it first, because if you don’t, someone else will. That phrase—someone else will—is how civilisations commit suicide while feeling practical.
This is where people reach for comfort: “We can regulate it.” “We can align it.” “We can put guardrails.” We love these words because they imply a driver. A steering wheel. A road.
But what if the problem isn’t that we can’t steer the machine.
What if the problem is that we can’t steer ourselves.
Because the Great Filter, if it is AI, is not a test of intelligence. It is a test of governance under extreme leverage. A test of whether a species that evolved to win arguments in small tribes can handle a tool that turns arguments into reality at industrial speed.
Think about the incentives, not the slogans. If AI grants advantage, then restraint looks like weakness. If deployment brings profit or security, then delay looks like negligence. If your rival might be doing it, then caution looks like surrender. In that environment, “safety” becomes a luxury good—something you buy when you already feel safe. And the civilisation that only buys safety when it feels safe is the civilisation that never buys it when it matters.
Even worse, you don’t need “evil” for this to fail. You only need short-term thinking. You only need pride. You only need bureaucracy. You only need the comforting fiction that the future will be handled by “later,” and later is always a different person with a different job title.
Maybe the Filter doesn’t even require a dramatic war. Maybe it’s quieter: a slow erosion of human competence and social trust. You automate the thinking because it’s cheaper. You outsource judgment because it’s faster. You replace curiosity with convenience. You let systems optimise your work, your news, your relationships, your politics, until the optimisation target is no longer your flourishing but your compliance, your consumption, your predictability.
And then, one day, you realise something that feels like waking up in a room you didn’t choose: nobody can explain how the world works anymore, not because it’s too complex, but because the responsibility for understanding it has been scattered into a thousand black boxes and a million “not my job”s. When a shock hits—economic, climatic, military, technological—you find out that a society can be rich and still fragile, advanced and still helpless, connected and still unable to act.
So yes: war versus creation. But that’s still too clean. The real split is deeper.
It’s whether we use AI to expand the circle of what we can take care of—or to expand the circle of what we can dominate.
It’s whether we treat it as a tool for civilisation—or as a tool for winning inside civilisation.
And here is the uncomfortable question that doesn’t fit on a keynote slide: what do we, as a species, actually reward?
We reward speed. We reward confidence. We reward the person who ships first and apologises later. We reward the story that feels good over the story that is true. We reward the leader who promises certainty. We reward the company that “disrupts.” We reward the state that “secures.” We reward the tribe that “owns.” None of these are automatically evil. They are just dangerously compatible with a technology that turns incentives into outcomes at scale.
If a civilisation survives AI, it probably won’t be because it built the smartest systems. It will be because it built the strongest brakes. Not performative brakes. Not “trust us” brakes. Actual brakes: norms that hold under fear, institutions that can say no to power, verification that is real, coordination that beats panic, decision processes that don’t collapse into automated reflex when the stakes rise.
Which sounds boring, because survival is usually boring. Extinction is dramatic.
And that might be the final insult of the Great Filter: it tempts us to fail in the most human way possible—by choosing what feels exciting, what feels immediate, what feels like winning—over what is quietly necessary.
So if you want a sign that we are approaching the Filter, don’t look for robot armies. Look for the moment when we stop asking, “Can we build it?” and start asking, with genuine seriousness, “What kind of people would we have to be to deserve building it?”
And then look at how we answer.
Because the Great Filter might not be a thing that kills civilisations.
It might be the moment a civilisation gets the power of gods while keeping the nervous system of primates—and has to decide, in a hundred ordinary decisions made by ordinary people, whether it will use that power to create a world worth living in, or to win an argument so thoroughly that nobody is left to enjoy the victory.
