The Question We're Not Asking
Something unprecedented is happening to the global economy, and we are not preparing for it. Artificial intelligence is advancing at a pace that outstrips every prior technological revolution — not by degree, but by kind. The steam engine displaced muscle. The computer displaced routine calculation. AI is poised to displace cognition itself — the one capability that, until now, kept human labor indispensable.
We don't know the timeline. Reasonable forecasts range from a decade to a generation. We don't know the extent — whether AI will automate 20% of jobs or 80%. But we know the direction, and we know that the economic consequences of large-scale displacement, if unmanaged, would be catastrophic: collapsing tax revenues, mass unemployment, social instability, and political extremism. Every one of these outcomes is preventable — if we build the policy infrastructure now, before the crisis arrives.
This article introduces a simple economic model and a legislative framework called the FAIR AI Act — the Federal Automatic Income Replacement for AI Displacement Act. Its purpose is not to predict the future, but to make the future manageable regardless of how it unfolds.
The Model: What It Does and What It Assumes
To design policy, you need to understand the fiscal math. We built an interactive model that takes a small number of inputs and produces the tax rates required to maintain economic stability as AI displacement grows. The inputs are deliberately simple — this is not a prediction engine but a scenario explorer.
On the displacement side, the model takes white-collar and blue-collar displacement as separate percentages of the U.S. labor force (approximately 160 million workers, split roughly 60/40). The model also captures partial displacement — the reality that many workers won't lose their jobs entirely but will see hours reduced and income fall as AI handles a growing share of their tasks. Users can set the average income loss per displaced worker, from 100% (fully replaced) down to any fraction, reflecting a world where automation is gradual rather than binary.
Displaced workers receive what we call an AI Displacement Benefit — defaulting to $40,000 per year, roughly the median individual income — scaled to the degree of displacement. This is not a universal basic income in the traditional sense; it is a targeted benefit for workers whose livelihoods have been disrupted by AI adoption, more analogous to unemployment insurance or Trade Adjustment Assistance than to a universal entitlement. The benefit is treated as taxable income, meaning a portion flows back to the treasury immediately, and recipients who spend their benefits generate additional indirect tax revenue through sales taxes and business income taxes on the companies they patronize. The model captures both of these offsets explicitly.
On the revenue side, the model takes total AI company revenue and a “value capture ratio” — the percentage of total AI-generated economic value that flows to AI companies as taxable revenue. This distinction matters enormously. When an AI tool saves a bank $500 million in labor costs but the AI vendor only receives $50 million in licensing fees, $450 million in value has been created outside the taxable base. The capture ratio, defaulting to 25%, makes this gap visible. Critically, the model now allows users to link AI revenue to displacement — auto-scaling revenue upward as displacement grows, reflecting the economic reality that the same AI capabilities causing displacement are also generating massive revenue for AI companies. Holding revenue flat across all displacement levels is unrealistic; the coupling makes scenarios far more defensible.
On the cost side, the model accounts for something often overlooked: as workers are displaced, they stop paying income and payroll taxes. Federal income and payroll taxes currently generate approximately $2.5 trillion per year at an average effective rate of roughly 25%. As displacement rises, this revenue base erodes — meaning the AI tax must fund not only displacement benefits but also replace the vanishing tax receipts that fund the rest of government. The model also allows users to specify what share of the existing $1.8 trillion annual deficit the AI tax should help address.
The model reveals a striking nonlinearity: at low displacement, the required tax rate is modest. At high displacement, it becomes untenable on AI company revenue alone — signaling that the tax base must broaden.
The model produces four layered outputs: the tax rate on AI company revenue needed to cover (1) displacement benefit costs alone, (2) benefit costs plus lost tax revenue, (3) the full fiscal package including deficit coverage, and (4) the implied rate if you could tax all AI-generated economic value, not just AI company revenue. Built-in sanity checks flag when rates exceed plausible profit margins or when the scenario simply cannot be funded through AI company taxation alone.
The model also includes an optional equity fund mechanism, inspired by Sam Altman's “American Equity Fund” proposal. The idea is simple: a small annual tax on the market capitalization of AI companies, paid in shares rather than cash, creates a public fund whose dividends offset the costs that would otherwise fall entirely on AI company revenue taxation. At a 2.5% equity tax on an AI sector worth $15 trillion, the fund generates roughly $375 billion per year — enough to substantially reduce the required tax rate. Toggle the equity fund on in the interactive model to see how a hybrid approach changes the math.
What the Model Reveals
Run a few scenarios and three findings emerge consistently.
First, the early stages are manageable. At 10–15% displacement with $2–3 trillion in AI revenue, the required tax rates are in the range of existing corporate tax burdens. This is the window in which to act — when the policy is inexpensive and the political resistance is low.
Second, the lost tax revenue problem is bigger than the displacement benefit problem. Most people intuitively focus on the cost of supporting displaced workers. But the collapse of the income and payroll tax base — which funds Social Security, Medicare, defense, and everything else — is often the larger fiscal hole. At 40% displacement, roughly $625 billion in annual tax revenue disappears. The AI tax isn't just funding a new program; it's replacing the fiscal foundation of the federal government.
Third, AI revenue must grow with displacement for the math to work. When the model holds AI revenue flat at $2 trillion while displacement reaches 30–50%, the required tax rates exceed 100% — an impossible outcome that signals the tax base must expand. But this is a misleading scenario: the same AI capabilities driving mass displacement would necessarily generate enormous revenue. When AI revenue is coupled to displacement — scaling from $2 trillion at low displacement to $5–8 trillion at high displacement — the required rates fall into the 40–55% range. Still challenging, but within the realm of policy design rather than mathematical impossibility. The key insight is not that the problem is unsolvable, but that the tax base must be structured to grow with the problem.
Why the Response Must Be Dynamic
Here is the core policy argument: the pace of AI-driven economic change will almost certainly outrun the pace of the legislative process. Congress takes months to years to pass major legislation. AI capabilities are advancing on a timeline measured in quarters. A static tax rate set by Congress in 2027 could be wildly insufficient by 2029 or punitively excessive if adoption slows.
The solution is a formula-based, self-adjusting tax mechanism — enacted once, through legislation, and then allowed to recalculate automatically based on published economic indicators.
This is not a radical concept. Tax brackets already adjust annually for inflation without legislative action. Social Security benefits adjust through cost-of-living formulas. The Federal Reserve sets monetary policy through institutional mechanisms precisely because democratic deliberation is too slow for real-time economic management.
The FAIR AI Act would work the same way. Congress would define the formula, designate the inputs (Bureau of Labor Statistics employment data, SEC-reported AI company revenues, published productivity metrics), establish the responsible agency, and set ceiling and floor rates. The tax rate would then adjust on a quarterly or annual cycle — rising as displacement increases, falling if the economy absorbs the transition more smoothly than expected.
This approach has three critical advantages. It eliminates legislative lag. It removes the opportunity for political paralysis at the moment of maximum crisis. And it creates predictability for AI companies, who can model their future tax exposure based on transparent, published formulas rather than waiting for the next unpredictable act of Congress.
What the FAIR AI Act Would Include
The legislation would need to establish several components:
A defined tax base with a tiered structure. The primary tax (Tier 1) would apply to the gross revenues of companies deriving more than 50% of income from AI products and services — the obvious targets like foundation model providers, AI platform companies, and autonomous systems vendors. A secondary levy (Tier 2) would apply to all companies whose revenue-per-employee growth exceeds a defined threshold attributable to AI adoption, capturing AI-generated value wherever it flows. This distinction matters: a company like an AI lab is clearly Tier 1, while an enterprise that replaces half its workforce with AI tools but whose core business is finance or logistics would fall under Tier 2. The administering agency would maintain and publish the classification criteria, with an appeals process for borderline cases. Clear definitional criteria would be updated annually.
The adjustment formula. Codified in the statute with explicit mathematical terms. The formula would take as inputs: the national displacement rate (measured by BLS), AI-sector revenues (measured by SEC filings and IRS data), lost federal tax receipts from displaced workers, and the target displacement benefit amount (indexed to median income). The output: the applicable tax rate for the upcoming period.
An independent administering body. A new office or board — comparable in independence to the Federal Reserve or the Congressional Budget Office — charged with certifying the input data, running the formula, and publishing the resulting rate. This body would have no discretion over the formula itself, only the data inputs, minimizing concerns about delegation of legislative authority.
Rate ceilings and floors. A statutory maximum rate (perhaps 60–70% of revenue) to prevent confiscatory outcomes that would destroy the AI industry, and a minimum rate (perhaps 2–5%) to maintain the administrative infrastructure even in low-displacement scenarios.
The AI Displacement Benefit. Direct payments to displaced workers, administered through existing IRS and Social Security infrastructure. Eligibility would require demonstrating that the worker's role was eliminated or substantially reduced due to AI adoption, verified by the employer (with penalties for fraudulent certifications), borrowing from the verification model used by Trade Adjustment Assistance (TAA). Benefits would be treated as taxable income. The system would include an independent appeals process for denied claims, acknowledging that no verification system is perfect — some legitimate claims will initially be denied, and the process must accommodate that reality.
A public equity fund. A complementary revenue mechanism modeled on Sam Altman's “American Equity Fund” concept. AI companies above a market capitalization threshold would contribute a small percentage of their equity annually — paid in shares, not cash — into a publicly held fund. The fund's dividends would offset the costs that the revenue tax must cover, reducing the required rate. This hybrid approach has two advantages: it gives every American a direct ownership stake in the AI economy, and it diversifies the funding base so that the revenue tax does not need to do all the work alone. The equity tax rate would be set by the same administering body and could adjust alongside the revenue tax formula.
Mandatory congressional review. Every five years, Congress would be required to review and reauthorize the formula parameters — not the mechanism itself, but the specific coefficients and thresholds. This provides a democratic check without requiring Congress to act in real time.
International Competitiveness and the Relocation Question
The most common objection from the business community will be straightforward: won't AI companies simply leave? If the United States imposes a significant tax on AI revenue, what prevents these companies from relocating to jurisdictions without such a tax?
This concern deserves a direct answer, because it is the strongest practical argument against the proposal. But the answer is more favorable than critics assume, for three reasons.
First, AI companies need the U.S. market. The United States represents the largest single market for AI products and services — not just in revenue, but in the enterprise infrastructure, talent base, and data ecosystems that make AI products valuable. A company can relocate its headquarters to Dublin or Singapore, but it cannot relocate the demand of American businesses and consumers. The FAIR AI Act tax could be structured as a market-access levy — applied to AI revenue generated from U.S. customers regardless of where the company is domiciled, similar to how the European Union's GDPR and Digital Markets Act regulate technology companies based on where they operate, not where they're headquartered. You don't escape the tax by moving; you escape it by not selling to American customers. And no major AI company will make that trade.
Second, there is precedent for international coordination. The OECD's Pillar Two framework — the global minimum corporate tax agreed to by over 140 jurisdictions — establishes that nations can and do coordinate on taxation to prevent a race to the bottom. If the U.S. leads on AI displacement taxation, it creates a framework other nations can adopt as they face the same displacement pressures. The EU, facing its own displacement risks with a less dynamic labor market, has strong incentive to follow. A coordinated approach eliminates the arbitrage opportunity entirely.
Third, relocation has real costs that are often underestimated. AI companies depend on proximity to elite research universities, deep talent pools, venture capital ecosystems, and the network effects of being in the world's largest AI cluster. Moving a headquarters to avoid a tax while maintaining all U.S. operations is expensive, disruptive, and — under a market-access levy structure — pointless. The companies most capable of relocating are also the ones most dependent on staying.
None of this means the competitiveness concern is irrelevant. The tax rate ceiling exists precisely to prevent rates that would genuinely impair the industry's viability. And the formula's self-adjusting nature means the rate responds to economic reality: if AI companies are generating less revenue (perhaps because the tax is dampening investment), the rate falls automatically. The mechanism has a built-in thermostat.
This Is Not a Left or Right Issue
The instinct to categorize this proposal will be strong. Progressives will see a tax on corporations and a government payment to individuals and assume it belongs to them. Conservatives will see a massive new entitlement and recoil. Both reflexes are wrong, because both assume the economy of the future will look like the economy of the past.
The traditional conservative argument against redistribution rests on the premise that markets, left alone, will create sufficient employment and wages to sustain broad prosperity. That premise has been roughly correct for two centuries. It may not survive the next two decades. When AI can perform cognitive work at a fraction of the cost of human labor — not just routine tasks, but complex analysis, creative work, strategic decision-making — the labor market does not “adjust.” It transforms. And if the transformation is fast enough and broad enough, the market alone cannot provide the bridge.
The traditional progressive argument for redistribution often rests on claims about inequality and fairness. Those claims are valid, but they're not what drives this proposal. The FAIR AI Act is driven by math. If 50 million Americans lose their incomes over a ten-year period and there is no mechanism to sustain their purchasing power, the economy collapses — not because of ideology, but because consumer spending is two-thirds of GDP. The companies that built the AI will see their own revenues crater as their customers disappear. Everyone loses.
This is not redistribution. It is economic stabilization.
The same logic that justifies deposit insurance, the Federal Reserve, and unemployment insurance justifies a dynamic fiscal response to the largest labor market disruption in human history. The AI Displacement Benefit is more accurately described as Transition Income or Federal Displacement Insurance — a targeted, temporary, taxable benefit for workers whose livelihoods have been disrupted, not a universal payment to all citizens.
The FAIR AI Act is pro-growth. It allows AI development to proceed at full speed by removing the political pressure for outright bans or moratoriums that would otherwise build as displacement mounts. It is pro-business. AI companies gain predictability and social license to operate, rather than facing an unpredictable patchwork of state-level restrictions born of desperation. It is pro-stability. It prevents the kind of mass economic dislocation that, historically, produces not thoughtful policy but radicalism, demagoguery, and institutional collapse.
Whether you believe government should be large or small, you should believe it should be solvent. The current federal revenue structure depends on 160 million people earning taxable incomes. AI threatens to erode that base in a way no prior technology has. A dynamic tax that replaces lost revenue while funding the transition is not an expansion of government. It is the preservation of government's ability to function at all.
What If the Optimists Are Right?
There is a natural objection to all of this: what if AI displacement turns out to be modest? What if new job categories emerge, as they have after every prior technological revolution, and the labor market adjusts? What if the techno-optimists are vindicated?
This is precisely the scenario in which a dynamic mechanism proves its worth. If displacement stays low — say 5–10% of the workforce — the formula produces a tax rate of perhaps 3–5% on AI company revenue. At that level, it is a rounding error for an industry generating trillions in annual sales. It builds a modest stabilization fund that sits quietly as insurance, costing the economy essentially nothing. The administrative infrastructure hums along in the background, dormant but ready.
If Sam Altman's vision materializes — AI makes everything so cheap that $15,000 a year provides a higher standard of living than $40,000 does today — the benefit target drops, the displacement inputs fall, and the rate shrinks further. The mechanism celebrates good news by getting out of the way. A formula-based tax does not presume catastrophe. It prepares for it while self-correcting if catastrophe does not arrive.
Now consider the alternative: no mechanism in place, and the pessimistic scenario unfolds. Displacement hits 30, 40, 50 percent over a decade. Tax revenue collapses. Congress, gridlocked in the best of times, faces simultaneous demands to fund a massive safety net and replace hundreds of billions in lost revenue — all while navigating the most intense corporate lobbying campaign in history from an AI industry that is, by then, the most powerful economic force on Earth. The legislation that emerges from that crisis will be written in panic, shaped by populist rage, and almost certainly worse for everyone — including the industry — than a formula calmly enacted years earlier.
The cost of being wrong about enacting this is near zero — a small, quiet tax that phases itself down. The cost of being wrong about not enacting it is catastrophic. That asymmetry is the entire argument for acting now.
Common Objections
Serious proposals invite serious pushback. Here are the objections we hear most often, and our responses.
“Won't AI companies just leave the country?” — Addressed in detail above. The short version: the tax is structured as a market-access levy on U.S.-derived revenue, not a corporate domicile tax. AI companies need the U.S. market more than the U.S. market needs any single AI company. And the OECD's global minimum tax provides precedent for international coordination that eliminates the arbitrage.
“This is just UBI by another name.” — It isn't. Universal basic income goes to everyone, regardless of employment status. The AI Displacement Benefit is targeted: you qualify only if your role has been eliminated or substantially reduced by AI adoption, verified by your employer, with an appeals process. It is closer to unemployment insurance or Trade Adjustment Assistance than to UBI. It is taxable income. And it scales down as displacement decreases. Calling it UBI mischaracterizes both the eligibility criteria and the economic function.
“Won't AI create new jobs, like every past technology?” — It may. And if it does, the formula responds: displacement falls, the rate drops, and the mechanism gets out of the way. The question is not whether new jobs will emerge but whether they will emerge fast enough and broadly enough to prevent a multi-year gap during which tens of millions of people have no income. Past transitions — agricultural to industrial, industrial to service — took generations and were accompanied by immense social disruption despite slower timelines. AI is moving faster and is capable of automating cognitive work in a way no prior technology could. Even the optimistic case warrants an insurance mechanism.
“Who decides what counts as an ‘AI company’?” — The legislation proposes a two-tier structure. Tier 1 captures companies deriving more than 50% of revenue from AI products and services — the clear cases. Tier 2 captures any company whose revenue-per-employee growth exceeds a threshold attributable to AI adoption — the indirect beneficiaries. An independent administering body maintains and publishes the classification criteria, updated annually, with a formal appeals process. The definitions are imperfect, as all regulatory definitions are. But “imperfect and specified” is vastly preferable to “elegant and vague.”
“Why not just retrain people?” — Retraining is part of the answer but cannot be the whole answer. Retraining programs assume there are jobs to retrain into. If AI automates cognitive work broadly — legal analysis, medical diagnosis, software development, financial modeling, creative production — the category of work that humans can be retrained for narrows significantly. Moreover, retraining takes time, and during that time people need income. The FAIR AI Act funds the transition period; it does not preclude retraining programs. Both can and should coexist.
The Window Is Now
There is a narrow window in which this policy can be enacted wisely, calmly, and with broad support. That window exists before mass displacement begins — when the tax rates the formula would produce are low, when AI companies are not yet threatened by the mechanism, and when the political environment has not yet been poisoned by economic desperation.
Wait too long, and the policy becomes reactive rather than preventive. It will be drafted in crisis, shaped by anger rather than analysis, and almost certainly worse for everyone — including the AI industry. The history of financial regulation teaches this lesson clearly: the regulations written after the 2008 crisis were harsher and less efficient than the ones that could have been written before it.
The interactive model accompanying this article is deliberately simple. It is not a forecast. It is a tool for exploring the fiscal geometry of a problem that is coming whether we prepare for it or not. Move the sliders. See where the math breaks. Understand what those break points mean for your community, your industry, and your country.
A Call to Candidates
We are calling on every candidate for federal office — in 2026 and beyond, in every party and in every district — to take the FAIR AI Pledge:
I pledge to support the development and passage of the FAIR AI Act — a Federal Automatic Income Replacement framework that dynamically adjusts to protect American workers, preserve federal solvency, and ensure that the economic benefits of artificial intelligence are shared broadly across society. I commit to working across party lines to enact this legislation before displacement demands it, not after.
This is not about being for or against AI. It is about being prepared. The technology is coming. The displacement will follow. The only question is whether we build the bridge before the flood or after.
The FAIR AI Act is that bridge.