UK shoppers face new taxes if AI replaces workers
Key Points
- The OBR models AI cutting the UK wage share from 40% to 20% of GDP by 2075-76
- Tax receipts fall nearly 4 percentage points of GDP in the scenario, even with total GDP unchanged
- Wages carry a 46% effective tax rate against 27% on profits, so the Treasury loses when AI shifts income to capital
- The OBR says new taxes on consumption, wealth, or AI use itself may be needed to fill the gap
- Around 10% of the UK workforce faces substitution by AI over the next decade, with 30% complemented
The Office for Budget Responsibility (OBR) warns that AI replacing UK workers could cut tax receipts by nearly 4% of GDP and force new taxes on consumption, wealth, or AI itself.
The warning appears in the watchdog’s Fiscal risks and sustainability report for July. The OBR models a scenario in which AI substitutes for human labour at scale over the next 50 years, shrinking the share of national income paid as wages and salaries from 40% of GDP in 2030-31 to 20% by 2075-76.
The OBR stresses the scenario is illustrative and sits at the upper end of estimates in the economic literature, but says much of that literature supports AI putting downward pressure on the share of income going to workers.
The problem for the Treasury is that the tax system leans heavily on wages. The OBR calculates that wages and salaries carry an average effective tax rate of 46% across its projection, while corporate profits carry just 27%.
If AI shifts income from workers to the owners of AI capital, the state collects far less tax on every pound the economy produces, even if the economy itself keeps growing.
Tax take falls even with GDP unchanged
In the OBR’s reduced labour share scenario, total GDP matches the baseline throughout, yet tax receipts fall from roughly 38% of GDP to under 35% by 2075-76. That gap of nearly four percentage points equates to well over £100 billion a year in today’s money.
The fall in taxes paid on wages is only partly offset by a rise in taxes paid on profits, because of the gulf between the two effective rates.
| Scenario | Wage share of GDP by 2075-76 | Tax receipts as share of GDP | Notes |
|---|---|---|---|
| Baseline | 40% | Roughly 38% | Assumes the wage share holds at its 2030-31 level, in line with 40 years of stability through the ICT revolution |
| Reduced labour share | 20% | Under 35% | Total GDP unchanged from baseline; assumes effective tax rates on wages (46%) and profits (27%) stay fixed |
The OBR said filling that hole may prove difficult with the existing tax system, because additional corporate profits from AI may escape efficient taxation using current instruments.
It notes that some commentators have suggested increased taxes on consumption or wealth, or entirely new taxes levied directly on AI use, could be needed.
Any of those routes would land on UK consumers, either at the till through higher consumption taxes or through levies on the AI services households and businesses increasingly rely on.
1 in 10 UK workers exposed
The OBR’s own estimates, produced for its earlier Forecasting productivity briefing paper, suggest around 10% of the UK labour force could face substitution by AI over the next 10 years, with a further 30% having their work complemented by the technology.
The watchdog’s central estimate has AI adding 0.2 percentage points to annual UK productivity growth over the next decade, with scenarios ranging from under 0.1 to as high as 0.8 percentage points.
The report also flags wider fiscal risks from AI beyond the wage share.
These include AI profits accruing offshore or in parts of the economy that are hard to tax, higher public spending on structurally higher unemployment and retraining, and other countries building dominance in AI infrastructure and drawing capital away from the UK, pushing up interest rates.
The OBR adds that AI also offers opportunities, including cheaper delivery of public services that could ease pressure on spending.
The OBR’s baseline still assumes nothing changes, with the wage share holding steady as it has for four decades.
A lasting shift would require generative AI to behave fundamentally differently from previous general purpose technologies, which destroyed jobs but ultimately raised productivity and wages while leaving the labour share broadly intact.
