IMF: Re-thinking taxation for the AI era

On 14 October 2026, at the IMF annual meetings, a forum discussion looked at Rethinking taxation for the AI era. The panel consisted of Adrian Brown, the Founder and CEO at Windfall Trust; Ioana Marinescu, Professor at the University of Pennsylvania School of Social Policy and Practice; Miranda Stewart, Professor of Tax Law at NYU; and the moderator Ruud De Mooij, Deputy Director at the IMF Fiscal Affairs Department.

Artificial intelligence is reshaping economies, creating new opportunities for growth while raising questions about job displacement, rising inequality and strains on public finances. These shifts in the economic landscape raise pressing questions about how tax systems should adapt. Issues arise as to how and where profit from large tech companies should be taxed. There are also questions about labour versus capital taxation. Some consider that governments should impose special levies on AI such as levies on compute, chips, robots, data centres, energy or AI subscriptions. The forum explored how taxation could raise revenue and address distributional concerns while preserving innovation, technology diffusion and long-term fiscal sustainability.

It is possible that AI could grow the economy but tax revenue could decline, because taxation is collected to a large extent from labour, and employment may be disrupted by AI. In the OECD countries, around 50% of revenue comes from taxes on labour, so if labour is under pressure the tax revenue will also be under pressure.

Research has indicated that a 30% labour disruption over ten years could decrease government revenue by 3%-4%, and including the offshore effect this could be an 8% decrease. With the increasing welfare bill, the effect could be amplified further. Governments should be testing now what will happen to public finances and do the necessary fiscal stress tests.

Ioanna Marinescu noted that taxes on labour are less important in developing countries as a source of revenue. Developing countries rely more on consumption taxes for their revenue. The impact of AI on the labour force may not be so severe in developing countries, because fewer employees are in jobs like IT and finance that will be replaced by AI. On the other hand, developing countries may be less exposed to the positive impacts of AI.

Professor Stewart noted that revenue collection could be helped by AI, as it could improve compliance. On the other hand, AI could help taxpayers to optimise loopholes. Some tax administrations are still not using computers, and most are not yet using AI, but already some administrations are using AI to improve efficiency. There is still a lot of scope to use AI to improve audit by better analysis of data.
Businesses are increasingly using AI, and this helps their tax compliance. There are advantages and risks for business and for the tax administration. The tax system needs to be adjusted to avoid the potential future revenue shortfall. We could tax AI inputs, chips or compute, and that would give us more AI capability.

Adrian Brown considered that countries need to look at the balance between taxation of capital and labour. Any system taxing labour at a higher rate than capital is accelerating the trend towards automation; and this may not lead to the optimal outcome for labour. There are other taxes available for raising government revenue, including wealth tax, consumption taxes and inheritance tax. These can be used to maintain revenue during the transition to a more balanced tax system.

Professor Stewart noted that it is worth persisting with income tax and asset taxes. Corporate tax remains an important part of the government toolkit, especially in developing countries where income tax and social security contributions do not provide a large share of government revenue. International tax coordination is more important because of AI, as states must tax the largest companies more efficiently. If this is not done, we are giving up a part of the tax base. The corporate tax base could be redefined, as the OECD Pillar 1 rules attempted to do. The UN Framework Convention is another opportunity to redefine the corporate tax base. Although the proposed changes may be imperfect, we must move in that direction.