Recent research from a macro analysis team highlights that the rapid adoption and deep penetration of AI technologies are posing unprecedented challenges to modern tax systems.
Looking back at previous industrial revolutions, the first three technological transformations successively drove the establishment and refinement of personal income tax, corporate income tax, and cross-border consumption tax administration frameworks.
The response cycle of tax system reforms has also shortened from over eighty years initially to around ten years.
The core impact of AI falls on the labor tax base, as job contraction and income divergence among workers directly undermine the stability of labor-related tax revenues.
Through a comprehensive assessment using four indicators—AI exposure, labor share, labor tax dependence, and fiscal vulnerability—Germany, the United States, and Japan rank highest in tax system risk exposure, while emerging economies face relatively smaller impacts.
Against the backdrop of globalization and major power industrial competition, improving rules within the existing tax framework may be a more feasible response at this stage.
However, it cannot be entirely ruled out that the US and Europe, facing greater fiscal pressure, might push for the creation of new tax categories.
How Have Past Technological Advances Affected Taxation?
If AI is viewed as a new wave of transformative general-purpose technology, then the experience of previous technological revolutions in reshaping tax systems serves as an important historical reference for understanding current changes.
Labor, capital, and consumption form the three major tax bases of modern fiscal systems.
The first industrial revolution expanded the potential tax base of personal income, with the UK pioneering personal income tax in 1842, followed by Austria (1849), Italy (1864), and Japan (1887).
The second industrial revolution broadened the potential tax base of corporate income, with Japan independently levying taxes on corporate income in 1899, and the US introducing corporate income tax in 1909 under the name "business privilege excise tax," after which wars accelerated the global spread of corporate income tax.
The internet revolution drove continuous improvements in consumption tax administration, with the 1998 Ottawa Conference forming a global multilateral coordination framework, the 2013 G20/OECD launch of the 15 BEPS action plans, and the 2021 BEPS Inclusive Framework reaching consensus on the two-pillar international tax reform among over 130 jurisdictions.
How Does AI Impact the Three Major Tax Bases?
AI's impact on the labor tax base is mainly reflected in job contraction and wage divergence.
Acemoglu & Restrepo (2020, 2022) estimate that each additional robot per thousand workers in the US corresponds to a 0.18-0.34 percentage point decline in the employment-to-population ratio and a 0.25%-0.5% wage decrease.
Challenger, Gray & Christmas data show that from January to August this year, over 100,000 layoff announcements in the US mentioned AI factors, accounting for more than 20%.
AI may also weaken the taxability of the capital tax base.
The computing layer of the AI industry chain faces depreciation rule disputes, the data layer faces measurement difficulties, the model layer faces pricing challenges, and the application layer has gaps in tax source identification, distinction between services and operating profits, withholding obligation allocation, and consumption destination taxation.
Additionally, the non-neutrality of the current tax system may amplify AI's impact on taxation.
The consumption tax base will also face structural weakening brought by AI.
Wages are one of the main sources of household consumption funding, and AI compresses household income through both job contraction and wage pressure, indirectly affecting the consumption tax base.
Furthermore, AI supports consumers in self-producing content to replace external procurement, compounded by subscription sharing, prepaid pay-per-use consumption, and multi-layer API transfer calls, amplifying issues of tax base omission, measurement difficulties, and tax timing mismatches.
Which Countries Are More Vulnerable to Tax Impacts?
Focusing on the labor tax base, the report constructs a comprehensive risk exposure measured across four dimensions: employment share in high AI-exposure occupations, labor share, labor tax dependence, and fiscal vulnerability.
From the AI exposure perspective, developed economies combine higher AI readiness with higher shares of exposed positions, facing greater potential substitution risk to their labor tax bases; emerging markets have lower AI readiness, with generally lower employment shares in high AI-exposure positions.
From the labor share and labor tax dependence perspective, Western and Northern European developed economies (Germany, France, Austria, Finland, Spain, etc.) have labor compensation exceeding 55% of GDP, with more than half of total tax revenue coming from labor-related taxes.
From the fiscal vulnerability perspective, the top five countries in the sample ranking are Japan, Greece, Italy, the US, and France, with Japan's pressure mainly stemming from extremely high broad debt burden ratios and old-age dependency ratios.
Overall, Germany, the US, and Japan have the highest comprehensive risk exposure, while Sweden, Finland, France, and Canada are in a relatively higher range, and emerging market countries face significantly smaller comprehensive impact intensity.
How to Respond to AI's Impact on Taxation?
Facing the potential tax impact of AI technology, two mainstream approaches have emerged in academic and policy circles regarding tax system adjustment paths.
First, introducing new tax categories (robot tax).
Guerreiro, Rebelo, and Teles (2021) point out that taxing robots is the optimal policy when workers with fixed skills who are difficult to reassign remain employed; Thuemmel (2019) proves that the optimal robot tax rate in the US is positive.
However, opponents argue that robot taxes may inhibit innovation, and robots themselves do not have taxpayer status.
Second, optimizing rules within the existing framework.
The EU Council Directive (EU) 2022/2523 adopted on December 14, 2022, established a 15% global minimum effective tax rate, applicable in major jurisdictions from 2024, constituting indirect taxation on the profits of large multinational enterprises (including digital and AI companies).
The research team points out that rule optimization within the existing framework is more aligned with current reality.
From the first industrial revolution to the UK pioneering income tax in 1842, the second industrial revolution to Japan independently levying corporate income tax in 1899, and the internet revolution to the 1998 Ottawa Conference, the lag time for tax system responses has shortened from over 80 years to about 30 years, and then to about 10 years.
However, the trigger for tax system reform is not purely driven by technological shocks.
The key lies in whether AI triggers a fundamental shift in the social wealth distribution pattern.
Looking back at history, the center of social wealth shifted from landowners to wage workers, and then to corporate capital income, expanding the tax base foundation for new taxes such as income tax and corporate income tax; but the implementation of new taxes often still requires external catalysts such as fiscal pressure.
A single country introducing a brand-new structural tax targeting AI could easily trigger cross-border tax base outflow, but it cannot be entirely ruled out that the US and Europe, facing greater fiscal pressure, might push for new tax categories, and globally unified reform also faces significant resistance.
Currently, promoting tax system neutrality, optimizing income structure, reducing fiscal reliance on a single wage total, and participating in rule-making for data elements and cross-border digital services—these rule optimizations within the existing framework—are more aligned with current reality.
This article is based on a September 24, 2026 research report titled "How Does AI Impact Taxation?—AI and Macro Series Part Two," authored by Zhang Jingjing and Chen Ying.
Comments