Who pays for the country?

tl;dr

The country looks after us partly because it needs us. Our work and our income tax are a big chunk of our say in how things go.
If AI ends up doing most of the work, that say moves to whoever owns the AI. For Australia, that's mostly someone else.
Norway hit a version of this with oil and the fix was to own the thing doing the work. Our super is the closest thing we've got, as long as we act like owners.

I stopped reading my payslip properly around 2016. The tax lines were too depressing, so now I just check the money landed in my bank and move on (as you do).

Lately I've been thinking about those lines differently. That tax is a big part of why anyone in Canberra has to care what I think, and I'm not sure how long that lasts.

A few months back I was on an AI panel at an angel network evening in Sydney. We covered the usual: which jobs go first, how fast, whether we all end up on a universal basic income. Nobody asked who pays for the country once the work moves to AI, and I've been chewing on it since.

The jobs frame

Most of the AI conversation happens at the level of jobs, and Treasury is no exception. The 2026 Intergenerational Report names AI as one of five big transitions for the economy out to 2065 and keeps the long term productivity assumption at 1.2% a year, noting that AI will likely give productivity a boost. The hopeful read is simple: AI makes workers more productive, they earn more and pay more tax, and the budget looks better.

That's a fair base case. What it doesn't cover is the version where the AI IS the worker. Then the question stops being "what happens to my job?" and becomes "what does the country need me for?"

Why the country listens

Partly because it's the right thing to do, and partly because it needs us. Personal income tax on its own is about half of all federal tax. Add the GST we pay when we spend our wages and the company tax on profits our work creates, and most of the budget traces back to people doing stuff.

Some countries don't work like this. Economists call them 'rentier states', which is a fancy way of saying the government acts like a landlord. It lives off rent from something it owns, usually oil in the ground, instead of off taxing its people. These countries tend to be [less democratic](https://www.cambridge.org/core/journals/world-politics/article/abs/does-oil-hinder-democracy/67665D8D240C8F43CD4A2DCB35894071 "World Politics: Ross (2001) and less answerable to their people than others with similar money. The government's income doesn't depend on its people being healthy, educated or productive. It depends on the oil coming out.

Luke Drago and Rudolf Laine applied it to AI and called it the intelligence curse:

When powerful actors create and implement general intelligence, they will lose their incentives to invest in people.

Same mechanism, just with data centres instead of oil fields.

A paper called Gradual Disempowerment gets to a similar place. The economy, the government and even culture stay roughly on our side because they can't run without us taking part. Take away the need for us and the reason to listen goes with it, without anyone planning or voting on it. The authors think it could end in an "effectively irreversible loss of human influence over crucial societal systems", which is about as serious as a sentence gets.

What I Learned

A lot of our power as citizens comes from the country needing our work. If that need goes away, the power probably goes with it, slowly and without anyone deciding it should.

The top of the treadmill

In The Codification Treadmill I argued that AI keeps soaking up expertise until it reaches the point where someone has to sign their name and own the outcome. I still think that's right. The problem is how few people stand there: a board, an auditor, the partner signing off on the deal, the doctor signing the script. That's thousands of people, not millions. Enough to keep humans in charge of the big decisions, and nowhere near enough to pay for a country.

The treadmill also eats from the bottom. Drago and Laine call this pyramid replacement: companies stop hiring juniors first, then trim the next level up, and keep going. A grad role that never got created doesn't show up as a job loss, so wages and the tax base thin out well before the jobs data looks scary.

Two seats

The way I've been thinking about it, each of us can sit in 2 seats.

Taxpayer: we do the work, earn wages and pay tax on them. Our say comes from being needed.
Owner: we own a slice of whatever's doing the work and get a share of what it earns. Our say comes from being a shareholder.

Most of us sit mostly in the taxpayer seat, and so far that's been enough. When the work moves to AI, that seat shrinks and the say moves over to the owner seat. The question for any country then becomes "how much of the thing doing the work do our people own?"

TODAY WHEN THE WORK MOVES TO AI ───── ───────────────────────── People do the work AI does the work ↓ ↓ Wages Profits for whoever owns the AI ↓ ↓ Income tax (~half of federal tax) Taxed mostly where the owners are ↓ ↓ The country needs its people The country needs the owners

When people do the work, the country needs people. When AI does the work, the country needs whoever owns the AI.

That's the part that worries me for Australia. Frontier AI is mostly American, with a few Chinese companies. We'd collect some company tax and GST, but most of the money ends up wherever the owners are. If we host the data centres and buy the services without owning much of the AI, we're the tenant, paying rent to landlords overseas. And bracket creep has already pushed personal income tax to half of Commonwealth tax receipts and the PBO expects that share to keep rising over the next decade, so we're leaning hardest on exactly the thing that shrinks first.

I do this for the love of writing — each subscriber costs me money. But please subscribe anyway and spread the word.

Norway's answer

Norway found oil around the same time as a bunch of countries that turned into landlord states. The difference is that it put the money into a fund owned by everyone, worth about NOK 3.8 million per registered person at the end of 2025. That's over half a million Australian dollars each(!). Norwegians ended up as owners of their oil economy as well as taxpayers in it.

We've built an accidental version of this. Our super system held AUD$4.77 trillion at 30 June 2026, a lot of it in global shares, so most of us already own tiny slivers of the US companies building AI. What I'm less sure about is whether we act like owners. An owner who doesn't use their rights has the money without the say.

What fixing it could look like

I don't think there's one fix, and I'd want the numbers done before backing any of them. But the options fall into 4 buckets, roughly from easiest to hardest:

Act like owners: our super funds could vote their shares together on safety and governance at the AI companies they hold, and we can check how our own fund votes. This one needs no new law.
Tax AI profits, not just wages: shift some of the load from wages towards profits. Australia already applies a 15% minimum tax on big multinationals, so there's a floor to build on. The catch is that AI profits get booked wherever the owners are.
Bargain on hosting: frontier data centres need land and energy, and we've got plenty of both. That's leverage to ask for something back while it lasts, like a revenue share, an equity stake, guaranteed local compute or safety commitments.
Own a bigger slice on purpose: a deliberate national stake in AI through the Future Fund or super, which is the Norway route. It's the most expensive option, and the one that most directly keeps us in the owner seat.

Each of these has a catch, and I'll dig into how they'd work, what they'd cost and where they break in part 2.

Where I could be wrong

My prediction is that this shows up in the budget before it shows up in the unemployment rate, with personal income tax growing slower than company profits for a few years running. If by 2032 personal income tax is still more than half of federal tax and wages are keeping up with profits, I've got this wrong, and I'll be very glad I did.

The weak spot is timing. If AI stalls well short of doing most of the work people get paid for, Treasury's base case holds and this post is a thought experiment. I don't think it'll stall, and either way these fixes are easy early and very hard late.

Maybe I should start reading those tax lines again. They're also my seat at the table, and I'd like us to still have one after the work moves on. Which seat do you reckon you'll be sitting in?

Disclaimer: Thoughts are my own and do not represent any other parties.