The most consequential near-term threat from artificial intelligence may not be rogue machines but a quieter danger to government budgets, according to a Bloomberg feature examining how AI could erode the tax base.
The concern is structural: developed economies fund themselves largely by taxing wages. In OECD countries, income taxes and social-security contributions make up roughly half of all tax revenue, while technology that replaces workers is taxed lightly and often subsidized. If AI displaces large numbers of high-earning employees or pushes them into lower-paying work, the worry is that income-tax receipts could fall just as demand for unemployment benefits, retraining and other social spending rises, squeezing public finances from both directions.
The figures being cited are striking, though they come with heavy caveats. A study by the Windfall Trust found that in a worst-case scenario, tax revenue in developed countries could fall about 15% from labor displacement alone, and by as much as 28% once higher social spending is factored in.
The US is especially exposed, since between 65% and 83% of federal revenue is collected on wages and salaries. Europe faces a parallel risk to its value-added tax, because if wage income shrinks, consumer spending, and the VAT levied on it, could weaken too. These are projections of a worst case, not forecasts, and they hinge on assumptions about displacement that remain deeply uncertain.
Policymakers and companies are floating a range of responses, none simple. Ideas include a tax on automation or “robots,” a levy on AI computing power or token usage, higher taxes on capital and corporate profits to rebalance away from labor, and redistributive mechanisms like universal basic income or “AI dividends.”
Notably, OpenAI’s own industrial-policy report suggested governments raise capital-gains or corporate taxes or explore taxes on automated labor, and Anthropic has floated a digital dividend funded by taxing the AI sector. Each faces the same practical obstacle: defining and administering the tax. As one tax expert noted, it is genuinely hard to identify and count a single “token” in a way that could anchor a levy, and taxes that fall too heavily on automation risk slowing the productivity gains AI promises.
Why the Fear May Be Overstated
A substantial body of economic opinion argues this alarm is premature. The historical pattern, from industrial machinery to computers and the internet, is that technology reshapes work rather than eliminating it, and the US labor share of income stayed within a fairly narrow band from the 1950s through 2023 despite enormous change.
Skeptics like the Tax Foundation and the Information Technology and Innovation Foundation contend that displaced workers generally move into new occupations, so the tax base erodes only if labor’s share of income falls dramatically and persistently, which they consider unlikely. Optimists, including Elon Musk, go further, arguing AI-driven productivity growth could raise revenue and shrink deficits. The US Congressional Budget Office has stayed carefully non-committal, allowing that permanent displacement could cut revenue while also noting AI may create new tasks and jobs that offset the loss.
Why It Matters Regardless
Even if the catastrophic scenarios never arrive, the debate points to a real vulnerability in how modern states are financed. Governments depend on current revenue streams, not long-run equilibria, so even a temporary lag, in which firms cut payrolls in anticipation of AI gains before new jobs and output materialize, could open a painful budget gap. That timing risk is the sharpest version of the problem: wage taxes can fall faster than the economy adjusts.
The deeper point is that tax systems built for an industrial age of human labor may need rethinking for an economy where value shifts toward capital and compute, a slow and politically fraught process. Whether or not AI proves as disruptive as the gloomiest projections suggest, the mismatch between a labor-based tax system and a capital-heavy AI economy is a question policymakers are only beginning to confront.
