WILL AI MAKE ‘INCOME TAX REDUNDANT WITHIN FIVE YEARS’?

Will AI make ‘income tax redundant within five years’?

Well, yesterday’s Telegraph article highlights this very claim… made by Monzo founder, Tom Blomfied.

Strip away the headline, and the Telegraph piece is really making three claims.

First, AI could trigger a jobs shock, especially in white-collar work.

Second, Britain is unusually exposed because it is so service-heavy.

Third, if payrolls shrink, the tax base built on wages starts to wobble.

That is more interesting than the headline claim that income tax will be “redundant within five years”.

The full article also shows that Blomfield’s tax claim rests on an even bolder technological claim.

He says current AI tools can already outperform humans in narrow domains and predicts they will become generalisable by the end of 2026.

That matters because the fiscal argument only moves at the speed of the capability argument beneath it.

If that timeline is wrong, then “income tax redundant within five years” is wrong by construction.

In my own piece, The (Tax) Rise of the Robots, I framed the problem not as the theatrical death of income tax, but as the hollowing-out of a tax system that depends heavily on human participation.

I also made the point that corporation tax does not automatically fill the gap in any neat way, especially if AI proves deflationary and firms can operate with fewer people, less physical presence and more mobile profit.

The OBR says income tax is expected to raise £329 billion in 2025-26 and NICs another £205.4 billion.

Together that is 43.4 per cent of all receipts.

A revenue base of that scale does not become literally redundant on startup timescales.

So, I would now put the point more carefully: income tax is not about to disappear, but the assumption that labour can remain the state’s unquestioned fiscal workhorse is perhaps starting to weaken.

Where the Telegraph article is strongest is in the specifically British vulnerability story.

It cites Adzuna data showing entry-level job adverts down 35 per cent since ChatGPT launched, points to Morgan Stanley’s warning that the UK may be especially exposed because services made up 81 per cent of output, and the chart on page 3 shows graduate hiring down across engineering, sales, accounting and finance, scientific and QA, trade and construction, consultancy and retail.

That is the real pressure point.

The labour-tax state can be weakened long before whole professions vanish; it is enough to have fewer junior openings, thinner career ladders and slower wage growth.

Bank Underground has already pointed to research finding UK postings in AI-exposed occupations 5.5 per cent below their pre-ChatGPT trend by mid-2025, with the hit concentrated in junior roles, while the International Labour Organisation’s (“ILO”) broader view is that most jobs are more likely to be transformed than simply erased.

Fiscally, that still matters. PAYE can erode through attrition as well as apocalypse.

Blomfield suggests it could instead be replaced by a tax on the resources used to build and run AI. He said: “I don’t think we’ll tax human labour, we’ll tax compute, [meaning systems like] data centres, and then we will use the proceeds to pay for government.”

A “Compute Tax” might be rhetorically elegant, but policy-thin.

Compute is an input, not a clean tax base in the way wages are.

Tax it bluntly and you hit electricity use, capex and location choices. Tax it lightly and it does not raise anything like what labour taxes raise.

The Telegraph piece half-admits this in its final move, noting that even an AI-services tax would be difficult to implement and pointing to the political fragility of the UK’s 2% digital services tax.

In other words, the article itself contains its own rebuttal. Broadly that if governments struggled to sustain a relatively modest tech levy, a globally effective tax on compute or automated labour is not going to be simple.

Even OpenAI’s own paper is more careful than the Telegraph. It does not offer a single silver-bullet robot tax. It suggests modernising the tax base by leaning more on capital-based revenues, targeted measures on AI-driven returns and possible taxes related to automated labour, while pairing that with incentives for firms to retain, retrain and invest in workers. That is a more realistic frame than pretending one new levy on data centres replaces PAYE.

In my own article, I called this the “paradox of taxing progress”.

A badly designed ‘robot tax’ could punish productivity and innovation.

But a tax system that keeps loading costs onto human employment while making automation relatively lightly taxed is also a design choice, and one that looks increasingly to have a limited shelf life.

So, the question is not whether to freeze progress or cheer it on. It is how to tax the rents, gains and dislocations of progress without taxing useful innovation into the ground.

In short, the tax system will need a reboot.