Who gets to flip that switch — the Fable ban through the three forks of Civilization

One letter, one afternoon

On June 12, 2026, at 5:21 PM Eastern Time, Anthropic received a letter.

It came from the U.S. Department of Commerce, citing national security authority. The contents were an export control directive: stop any foreign national from using Claude Fable 5 and Mythos 5 — at home or abroad, including Anthropic's own foreign employees. To ensure compliance, Anthropic could do only one thing: shut these two models off foreverycustomer.

Within hours, a commercial product that had launched just three days earlier and was being used by hundreds of millions of people disappeared from the entire world.

What makes me uneasy isn't the question of "whether governments should regulate AI" itself. Dangerous things should have someone able to hit the brakes — I agree with that. What actually cut deep was themanner— a letter, an afternoon, a rationale that was only verbally explained and that even the developer itself said it couldn't pin down, was enough to take a globally deployed tool offline. Anthropic later said it had verified the alleged "jailbreak" technique, which amounted to nothing more than asking the model to read some code and find software flaws — a capability that exists just as much in other models and that the defensive side uses every single day.

In other words, what was pulled and the reason for pulling it may have nothing to do with each other.

What I want to talk about isn't whether this incident was right or wrong — there are legal processes to handle that. I want to talk about what itdemonstrated: everything you depend on has, at its very foundation, a switch you can't control or even predict. And the question of whose hand is on that switch is one the game thought through for us long ago.

Civ VI's Tier 4 governments

Anyone who played Civilization VI's "Gathering Storm" expansion will remember that, as the game reaches the end of the Information Era, three "Tier 4 governments" unlock. They aren't a better-or-worse relationship, butthree forks in the road：

• Synthetic Technocracy
• Corporate Libertarianism
• Digital Democracy

Back when I first played to this point, I was genuinely curious what these three roads would actually look like if they showed up in our world.

It's a brilliant design. It doesn't frame any one road as civilization's endpoint or correct answer. All three run on thesamedigital infrastructure — the same data, the same algorithms, the same connected world. The only difference is one thing:

who holds the power to define.

Technocracy puts it in the hands of algorithms and experts. Corporate libertarianism puts it in the hands of markets and capital. Digital democracy breaks it apart and spreads it across everyone who can speak. Three roads lead to three very different worlds, but they share the same foundation.

Keep one detail in mind: in Civ VI, these three governments aren't abstract labels — each is tied to a concrete set of rewards and costs. The designers used numbers to price each of the three forks. As you'll see, what those numbers say is more direct than any manifesto.

In the language of Civ, the Fable affair is a question of "which Tier 4 government you chose." And this time, we saw the first road at its most naked.

The first road: when the switch is held by discretion

Synthetic Technocracy in the game runs on centralized, top-down judgment. Its in-game effect is blunt: +3 Power to all cities, +30% Production toward all city projects — the most productive of the three roads, built for racing to a science victory or speeding up megaprojects like the Mars landing. But its "legacy cost" is −10% Tourism. In other words, this system bets everything on efficiency and output, and the price it pays is cultural appeal — the soft influence that makes others "want to be like you." It runs the fastest, but no one looks up to it anymore — and whether the system works at all hinges on whether you trust the hands holding the judgment.

The Fable ban is the real-world version of this road. A single act of administrative discretion, with no deliberation and no full disclosure of evidence, can take effect worldwide. The problem isn't "government intervention" — it's that themeans are plainly disproportionate to the stated end— pulling an entire model off the shelf globally in response to a verbally described vulnerability that exists elsewhere too.

This is exactly the core of my unease. When "whether you can use this tool" no longer depends on whether it's good or safe, but on the in-the-moment judgment of some administrative discretion, that sense of the ground shifting beneath you is real. Everything you've built, it turns out, was standing on borrowed ground.

This discontent shouldn't point in only one direction. Export controls aren't new this time around; they've always been running, just usually on chips, encryption, and other things you never see. The fact that AI models have been folded into this framework means, in a sense, they're finally being treated as "something with real strategic weight." It's unsettling precisely because it's effective enough to deserve being governed this way.

That doesn't make the letter's procedural problems legitimate. But it reminds us: frontier AI that's completely beyond the reach of any force may not be a more reassuring world either. The question was never "should there be a brake," but "who should hit the brake, and by what process."

The second road: when the switch is held by the market

So what if we leave the government out of it? Hand the switch to companies, to the market — wouldn't that be freer?

That's the temptation of the corporate libertarianism road. On the surface, there's no arbitrary government discretion; tools live or die by supply and demand. It sounds clean.

But its numbers in Civ VI are far more violent than the name. This government called "freedom" actually grantsthree military policy slots— the most war-and-expansion-oriented of the three roads, raising a large army on the production bonuses of commercial hubs and encampments. Its legacy cost is −10% Science. In other words, this "market-first" road actually grows its muscle in force and resource seizure, and the price it pays is long-term accumulation of knowledge. It maximizes immediate production and possession, sacrificing the ability to see further ahead.

But the Fable affair also exposed this road's hole card. Even without that letter, Fable's survival was always in Anthropic's hands — one company, one jurisdiction. Today it's the government flipping the switch; tomorrow it could be a business decision, a pricing strategy, some shift in priorities. My entire workflow rests on a single vendor's models at its foundation. This time it was Fable that got pulled, but the structural problem doesn't disappear by switching to market control.

Moving the switch from the government's hands to a company's doesn't remove the switch — it just changes whose hands it's in. And those hands are just as far beyond your reach.

The third road: what about digital democracy?

So, naturally, you think of the third road. If government discretion is unsettling and companies can't be trusted, then what about lettingeveryonetake part?

First, clear away a common misconception. The point of digital democracy isn't referendum-style vote-counting where "everyone casts a yes or no on a single issue." It's more like away of expressing views and forming participation— letting anyone read the arguments, put forward their own, respond, add to them, second them. Voting is just the most downstream, crudest move; the real core is scaling up the act of "speaking up," so that a good idea can come from anyone in the community.

The game's description of "digital democracy" means exactly this. It says: direct democracy's greatest virtue is that every citizen can speak on affairs of state in their own voice; but its fatal flaw is that it's extremely hard to scale — ever since the Athenians shouted themselves hoarse in the agora, political philosophers have searched for more scalable forms of democracy. Digital democracy's premise is to use technology to solve this scaling problem: issues are laid out for open discussion, comment, and contribution, and citizens can take part in this ongoing conversation right from their smartphones, rather than voting only once every few years.

Its numbers in Civ VI fit this character well: +2 Amenities in all cities, +2 Culture per specialty district, leaning toward culture and diplomacy victories by keeping people happy and spreading culture. The civic it unlocks is called "Distributed Sovereignty" — power broken apart and flattened. And its legacy cost is −3 Combat Strength: a system that spends its energy on participation and consensus is inherently weak at making a fist. It's the most enchanting, and the most toothless.

It sounds beautiful. But the game goes on to point out the part that should set off alarms — it says many online communitiesalreadyhave built-in mechanisms for measuring engagement: trending votes, likes, shares, which roughly estimate participants' social capital, and some believe this can be extended to political questions. The game even slips in a note of doubt: early optimists thought digital democracy could sidestep demagoguery and malicious manipulation, but that confidence should be held lightly — whether it ultimately delivers on the ideal of universal participation, or gets forever swept away by cat pictures and flame wars, remains unknown.

That's where the trap is. The problem isn't "participation" itself — letting more people speak up and add to the arguments is exactly the best part of digital democracy. The trap is that when "participation" degrades into likes, shares, and instant reactions, when social capital becomes something that can be metered and mobilized by the second, the slow kind of participation — the kind that requires reading others' arguments before speaking — gets crowded out by the fast kind that needs only an emotional reflex.

And a "digital democracy" that pressures decisions in real time through likes and shares is, with that letter that made me uneasy,structurally the same thing— both bypass the slow, the cumbersome, the checked-and-balanced process to reach a result that feels right in the moment.

Today you can use it to force the government to restore Fable. Tomorrow, the same mechanism can be used to force the government to pull something else.

So on the Fable affair, the stance that's trulyconsistentwith my unease is precisely not a call to mass protest, not real-time mobilization. Because what I'm objecting to in the first place is that "the process was skipped." If I oppose it using a means that skips the process just as much, then I'm not opposing the principle of the matter — only thetarget。

What should make me uneasy isn't "the process got in my way," but "the process got bypassed." The difference between those two kinds of unease is the difference between the two exits of the third fork: one leads to deeper deliberation, the other to a faster mob.

And there's also the fact that I've been bound by American sovereignty — when I'm not even an American citizen!

The foundation all three roads share

Back to Civ's design. Its most honest stroke is letting you see clearly that all three roadsshare the same foundation: vast data, real-time algorithms, a connected world. Technocrats, the market, the masses — they're just three different sets of hands reaching for the same switch.

The Fable ban gave us a rare, clear look at that switch's existence. Normally it stays well hidden. You open the app, type a prompt, get a response — the whole process so smooth you forget that, at the very bottom of all this, there's a switch that isn't in your hands.

My discontent with this ban is real, and I don't intend to digest it away. But I also want to put that discontent in the right place: it shouldn't become an urge to "find another set of hands more to my liking," but a harder and more worthwhile question —

Is there a way for that switch not to be held by any single set of hands?

None of the three forks gives a clean answer to this question. Technocracy hands it to experts, at the cost of no one wanting to be like you; companies hand it to the market, at the cost of being all muscle and no foresight; digital democracy tries to spread it across everyone, but easily lets the fastest, loudest emotion drown out the slowest, most worthwhile argument. For every road, the game honestly marked its cost. Maybe the real fourth road hasn't been walked yet; maybe it isn't in the question of "who should it be handed to" at all, but in the direction of "how to make it so no single party can flip it alone."

I don't have an answer yet. But I'm certain of one thing: before we've thought it through, rushing to wrest the switch from one set of hands into another never solves the problem of the switch itself.

That letter made me uneasy. But staying uneasy may be closer to the true weight of this matter than rushing to feel at ease.

Meaning comes from difference. So does a tool's survival — from whose power it's placed beside, and by what process it's weighed. Fable hasn't changed; what changed are the hands around it. And seeing those hands clearly is the first step toward thinking through the next one.

Further reading: - Fable and Mythos — what I was really thinking about when Fable 5 launched - From customer service to echo: in 74 days, I replaced the thing that spoke for me - It wants to be AI's upstream; I just want to leave an echo behind