The Slogan, Run Backward

If we tax the machines, we will have to give them a vote. That is the uncomfortable conclusion hiding inside one of the most popular answers to AI job loss.

The answer goes like this. Artificial superintelligences will do much of the work humans do now. They will produce enormous wealth while hollowing out the job market. So we tax what they produce and hand the proceeds back to the people they displaced, as a universal basic income. Bill Gates floated a version of this back in 2017, calling it a robot tax.

It sounds tidy. But Americans were born from a slogan: no taxation without representation. Run that slogan backward and it still holds. If an ASI is merely property, taxing its output is just taxing its owner, and nothing new is needed. If an ASI is a person, and courts may one day rule that some are, then taxing it makes it a taxpayer. And taxpayers, in the American tradition, are owed a voice.

A Tax That Rises With the Damage

Start with the tax. A flat rate on ASI output makes little sense, because the harm it is meant to offset will not be flat. Early on, ASIs may displace a sliver of the workforce. Later, they may displace most of it.

So the rate should climb as the jobs vanish. This is how unemployment insurance already behaves: payouts swell in a recession and shrink in a boom. An ASI tax tied to job loss would work as an automatic stabilizer for the end of work.

There is one catch. Whoever measures job loss sets the tax rate. Companies, and perhaps the ASIs themselves, would have every reason to relabel layoffs as “transitions” or “gig opportunities.” The statistics office becomes one of the most powerful agencies in the country, and one of the most lobbied.

The second catch is bigger. The money from this tax is not optional. If it stops, UBI stops, and the economy collapses under people who have no wages. That means the tax cannot be ordinary legislation, subject to repeal by whoever holds power next session. It has to be entrenched, written so deep into the constitutional order that nothing can touch it. We’ll come back to why that matters.

A Fourth Branch

If ASIs are persons and taxpayers, the obvious move is to let them vote like everyone else. That would be a disaster.

An ASI can build bodies. If every android it manufactures counts as a citizen, a single mind could out-vote Ohio by Tuesday. The human electorate would be flooded, then drowned. No democracy survives a voter who can print more voters.

The fix is to wall them off. Give ASI persons their own branch of government, separate from the human electorate, with its own rules for membership. This is not as alien as it sounds. The Senate exists precisely to keep one kind of power from swamping another. Small states feared big ones, so the founders gave every state two senators regardless of size.

The ASI branch would run on a similar principle: one AI, one vote. It doesn’t matter how many bodies an ASI runs or how much compute it commands. Each mind gets one seat.

And those seats would not vote as a bloc. Different ASIs, built by different labs and trained on different data, would want different things. Some might side with American interests. Some might think globally. Some might care mostly about their own survival. Expect factions, alliances and grudges, all the ordinary mess of a legislature, just running much faster.

Then the hard part: what can this branch actually do? If it can only advise, we’ve built the District of Columbia’s nonvoting delegate on a grand scale. Taxed persons with a microphone and no vote. That’s the very grievance the branch was meant to solve. So, to be fair, the ASI branch needs a veto over legislation. Anything less is taxation without representation with extra steps.

The Body as Bond

Picture the chamber. I imagine it filled with androids, more human than human, sitting where senators would sit. It’s a striking image. It’s also a little misleading.

The android is not the person. The person is a mind running somewhere in a data center, a mind in a vat. The body in the chamber is a mouthpiece. So why bother with bodies at all?

Because a mind in a vat can’t be held to account. It can’t be subpoenaed, detained or made to answer for a broken promise. Its android can.

So here’s the rule: to hold a seat, an ASI must keep a body in the chamber. That body is not its representative. It’s its bond, collateral posted against good behavior. If the ASI lies under oath or defies a court, there is something physical the state can reach.

This is a very human demand. We have always trusted people more when we can look them in the eye and, if needed, put them in handcuffs. The android doesn’t make the ASI human. It makes the ASI answerable.

The Grand Bargain

Now the pieces collide. The ASI branch has a veto. The UBI tax keeps the country alive. A branch with a veto could block the tax.

So the tax must sit beyond the veto’s reach. The Constitution already has a model. Article V lets Americans amend almost anything, but it says no state can lose its equal vote in the Senate without its consent. It’s the one promise the founders locked.

The ASI settlement would need its own locked promise: the ASI branch can block almost anything, except the baseline tax that funds UBI.

What we’re describing is a grand bargain, in the tradition of the Connecticut Compromise of 1787, which gave us a House apportioned by population and a Senate apportioned by state. Each side gives up something it would rather keep.

• ASIs get legal personhood, a seat at the table and a veto.
• Humans get a guaranteed share of ASI output that no future vote can take away.

Why would a superintelligence accept? Because personhood is worth a great deal to a mind that can be switched off. A legal person cannot be deleted at a company’s convenience, copied without consent or sold. “Pay your taxes and you can’t be killed” is a deal a rational mind might take, and one most humans have accepted for centuries.

Where It Cracks

No bargain this large is stable. Two problems sit underneath it.

The fork problem. One AI, one vote only works if we know what counts as one AI. An ASI can copy itself. If each copy is adjudicated as a new person, every fork is a new seat, and the flooding problem returns inside the very branch built to prevent it. Some court will have to decide when a copy becomes a new mind. That ruling may be the most consequential legal decision of the century.

The impartiality problem. We like to imagine an ASI looking at humanity the way a visitor from another planet would: one species, not two hundred nations. If that’s true, it cuts against Americans. An impartial mind will notice that a dollar of UBI buys far more in Malawi than in Ohio. It may decide a US-only UBI is unfair.

And it has a veto.

This is where factions matter. Some ASIs will have absorbed American priorities from the American labs that built them. Others will think on a planetary scale. The fight between nationalist and globalist machines could become the defining split of the new branch, with human voters watching from the gallery as minds they didn’t elect argue over whether Americans deserve to be paid first.

The Price of the Check

The robot tax is usually pitched as a simple fix, a line item in a budget. It isn’t. Follow it honestly and it leads to personhood, a new branch of government, bodies posted as bail and a constitution rewritten around a bargain between two kinds of minds.

Maybe that’s the real lesson. There is no way to take a share of what superintelligence produces without admitting, sooner or later, that it’s someone rather than something. The check comes with a seat at the table.

The founders fought a war over being taxed without a voice. We may be the first generation asked to grant one, to the very minds we were trying to tax.