RAND examines tax alternatives if AI shifts income from labor to capital
RAND compares business-income, profits-per-employee, capital-gains, consumption and wealth/inheritance options if AI shifts income away from wages.

Established trend · Active
Latest assessment: · Evidence updated
Researchers and policymakers increasingly ask what remains taxable if wage income or payroll contributions weaken: business profits, capital gains, consumption, tokens, compute, gross receipts or public ownership. The mechanism under debate is preserving government and social-insurance capacity when economic value migrates.
Scope: Serious fiscal-architecture research and concrete proposals; idea track distinguished from implementation.
Research and proposals examine different ways to locate taxable AI-related value and sustain public services. They have not established an adopted replacement fiscal architecture or an observed AI-caused migration of revenue. Tax bases, incidence and administrative feasibility remain central.
RAND compares business-income, profits-per-employee, capital-gains, consumption and wealth/inheritance options if AI shifts income away from wages.
Yale Budget Lab models federal revenue under alternative AI growth and income-share scenarios; CFR discussion adds tax-administration considerations.
Korinek/Lockwood explore taxation and public assets in hypothetical labor-displacement and autonomous-AI economies.
OECD explains how lower labor share could weaken social-insurance contributions even without net job loss.
Gates proposes comparable FICA treatment for human and robot labor to support pensions.
Casar’s H.R.10044 proposes an AI-usage excise base and worker/public-interest employment financing.
Illinois modeled data-center employment and fiscal effects: construction job-years differ from roughly 2,800 direct permanent jobs, and state net losses can coexist with local gains.
Later approved evidence — October5: Fiscal research connects AI tax bases and administrative scale; proposed cooperation is not a funded public institution.
Publisher contents identify the March 2026 article; an author/publisher-provided RePEc abstract compares taxes on autonomous-capital income, investment and employer robot social-insurance contributions under two substitution scenarios. The full publisher download failed.
This is serious fiscal research, not an adopted tax or an observed revenue effect. Full-paper details and calibration are held.
Lucas Irwin, Tung-Yu Wu and Fazl Barez’s position paper was first submitted March 4, revised June 11, and included in the July 6–11, 2026 ICML proceedings. June version of the paper ↗.
The proposal is a surcharge on model inference at the point of sale. A “token” is a unit of model input/output, not a currency token. The paper proposes compute-provider records and a staged audit system: estimate/check usage externally, use model-category norms if necessary, and inspect internal generation records as a backstop. It favors capture where services are used rather than only where a model is hosted.
The authors acknowledge that external token audits and norm methods are unsolved, that languages/models tokenize differently, and that self-hosted models complicate enforcement. They call for modeling who bears the tax, effects on activity and comparison with computation-based taxes. This is a research roadmap, not a demonstrated collection system, revenue estimate or adopted tax.
This is eligible for Lane 9’s fiscal-idea track because it specifies a taxable transaction and verification architecture. It extends the current Korinek–Lockwood discussion with implementation questions rather than merely repeating a general call to “tax AI.”
My inference is that delegating verification to cloud firms could build fiscal capacity while also strengthening private intermediaries. Public audit access and treatment of independent deployment would determine that balance.
A scheduled October 22–23 NBER Public Economics meeting includes Giacomo Brusco and Francesco Capozza’s AI-tax analysis. The accessible indexed abstract considers usage/token, profits, electricity and automation mechanisms. In its model, taxing use must balance distribution and externalities against a supplier’s existing restriction of output through market power. Some assumptions therefore imply little or no positive usage tax. These are model results, not measured revenue forecasts or a conclusion that every AI tax is harmful. The abstract does not provide the full model specification and parameter choices. The conference has not occurred, and the paper’s first publication date is unverified.
That is a meaningful Lane 9 idea: the tax base and the taxed firm’s market structure can change whether government captures surplus or further constrains access. It is distinct from current papers with different authors and mechanisms.
Ahmad Al-Yamani, Jude Althagafi and Leena Ghandourah argue that fragmented and AI-supported work can weaken institutions financed through stable employment. Their framework proposes portable protection funded by contributions prorated across engagements, contributions from entities engaging labor, and a broader shift toward value creation. Its one-worker-does-eight-workers’ output example illustrates potential payroll-base erosion; it is explicitly not a forecast. The work synthesizes conference discussions and existing research, not a new causal experiment. Takamol operates labor-market infrastructure and has an institutional stake in data-intensive coordination. The primary summary is accessible; the full design and initial publication metadata are not established. This is older work recirculated in October, with its exact initial publication date unresolved.
The important mechanism is a fiscal/social-insurance system that follows workers across jobs while capturing surplus elsewhere. Data integration also raises questions about oversight and workers’ rights to challenge records.
Emily Kraschel and Scott Moore published a Searchlight Institute analysis October 5 connecting planned data centers with rural Opportunity Zone incentives. Their mapping found 58 of 328 proposed projects in eligible rural tracts. Eligibility is not proof that a tract has been designated or that a project will receive tax benefits.
The fiscal mechanism is preferential treatment of investment, including a lower substantial-improvement threshold for developed property in rural zones. Undeveloped land receives different treatment. The authors discuss reforms to better connect incentives with public benefits. Their geographic overlap is not a causal finding: power, land and cooling requirements may also explain site selection.
For The Race, this raises the question of whether tax expenditures purchase benefits for communities or mainly reduce private costs. It belongs in the fiscal-idea track even without new legislation.
Distinct connection: For The Race, this raises the question of whether tax expenditures purchase benefits for communities or mainly reduce private costs. It belongs in the fiscal-idea track even without new legislation. Overlapping evidence is not independent confirmation.
Will Raderman proposes raising and indexing the taxable wage base of the Federal Unemployment Tax Act, rather than introducing a separate AI tax. Most covered employers currently pay a net federal rate of 0.6% on the first $7,000 of wages, normally $42 per employee.
Additional revenue would flow through a proposed Federal Adjustment Account within the Unemployment Trust Fund to help finance stronger state benefits. Federal tax credits would encourage states to meet improved benefit standards. The paper discusses several benefit options, including higher maximum payments and replacement rates reaching 75% of prior wages; those are policy designs, not enacted entitlements.
The argument draws on earlier Kennedy and Johnson reform proposals and existing federal-state financing machinery. Its revenue illustrations draw on other fiscal estimates, rather than an observed experiment or a demonstrated forecast of AI unemployment. Primary fiscal proposal ↗.
This belongs in Lane 9’s serious fiscal-idea track because it identifies a revenue base, an account, transfers and incentives. It strengthens social insurance, but remains tied to wages rather than directly capturing AI profits or rents.
Henrik Kleven and Owen Zidar's September 2026 paper, Automation and Optimal Taxation: A Task-Based Theory, asks how taxes and work subsidies should respond when machines can perform some occupations' tasks. It combines a theoretical model with US earnings and automation-exposure data. Full primary paper ↗.
In the model, automation changes wages differently across occupations. The authors find a case for stronger subsidies at the bottom, lower taxes on exposed middle earners and higher taxes on relatively protected upper earners. The work subsidy resembles an expanded refundable earned-income tax credit: eligible low earners can receive money through the tax system. The analysis does not imply that a special machine tax must rise merely because automation expands.
These are conditional welfare calculations, not observed results from current generative AI. The calibration uses pre-generative-AI earnings and exposure data; workers do not retrain or move between occupations in the model. Those assumptions are particularly important for interpreting its distributional results.
For The Race, the contribution is a concrete alternative to treating every displaced task as a separately taxable “robot”: existing labor-income and capital-income instruments can be redesigned around who loses wages and who receives income.
AI-capital taxation/transfer simulations identify a possible surplus base and fiscal limits; no tax adopted or empirical receipts measured.
A stylized wage-to-profit scenario identifies tax-base composition risk at unchanged GDP; no observed revenue loss or policy instrument.
This adds a relevant design question to our existing fiscal case: would a tax capture AI-related surplus, or alter which ownership structures can compete? The draft remains a proposal, with no verified new revenue estimate or enacted charge.
For The Race, the contribution is moving beyond an undifferentiated “robot tax” to locate collectible bases and public authority along the supply chain. A nominal tax on suppliers could still be passed to smaller users.
The comparison explains the administrative gap between AI activity and observable tax bases. That bears on governments’ ability to capture surplus, without establishing revenue yield or an adopted tax.
The proposed net-profit tax dedicates part of infrastructure returns to residents.
A cash-flow tax paired with a wage tax offers a way to tax excess returns; a separate wage insurance proposal addresses the transition for displaced workers.
Tax revenue varies with where surplus lands, including labor income, producer profits and consumers’ lower prices; the report does not predict the shift.
Unclear for fiscal authorities and social-insurance beneficiaries: the debate adds policy options, but no replacement revenue capacity is established.
Credible revenue and administration estimates, clear evidence of who pays and receives, and an adopted fiscal mechanism would show whether debate becomes governing capacity.