Should the People or the Power Fund Universal "High" Income?
AI displacement is accelerating. The policy debate about who pays for it has begun. Here is where each argument actually stands.
Amendment Media | Jim Pearl
On April 17, 2026, Elon Musk posted on X that “Universal HIGH INCOME via checks issued by the Federal government is the best way to deal with unemployment caused by AI.” The post drew 32 million views within hours. Sam Altman, CEO of OpenAI, has endorsed a similar concept. A New York state legislator running for Congress has called direct payments a necessary “insurance policy.” A UK government minister is weighing the same idea.
The question of whether artificial intelligence displacement should be met with universal income has moved from the fringes of policy debate to its center. What has not yet been settled — and what the debate largely avoids — is the more specific question underneath it: if a safety net for AI-displaced workers is necessary, who should fund it?
That question has more than one serious answer. This piece documents them.
The Scale of the Problem the Policy Is Trying to Solve
Before evaluating the funding proposals, it is worth establishing what they are responding to.
Nearly 78,600 tech workers were laid off in the first quarter of 2026 alone, with approximately 48 percent of those cuts attributed directly to AI and workflow automation. The World Economic Forum projects that 30 percent of current U.S. jobs could be automated by 2030, with 60 percent seeing significant task modification. Anthropic CEO Dario Amodei has warned that AI could eliminate half of all entry-level white-collar positions. Goldman Sachs has estimated displacement of 6 to 7 percent of the U.S. workforce.
These projections vary in their specifics and timeframes. They do not vary in their direction. The displacement is coming. The policy question is what, if anything, the government and private sector should do about it — and who should bear the cost.
Position 1: Government-Funded Universal Income (The Musk-Altman Proposal)
The most prominent proposal is the simplest to describe: the federal government issues regular payments to all Americans, or to displaced workers specifically, funded through general taxation.
Musk has described his version — which he calls “Universal HIGH Income” rather than UBI, suggesting more generous payments than traditional basic income proposals — as the most practical response to what he sees as inevitable mass unemployment. He argues that AI and robotics will produce goods and services far in excess of any inflationary effect, making the payments fiscally sustainable. Altman has funded experimental pilots through his OpenResearch project and argues that direct payments allow individuals to navigate displacement on their own terms rather than being funneled into government-directed retraining programs.
The experimental evidence from pilot programs is cautiously encouraging on one specific question. Altman’s own pilots, along with trials in Stockton, California and elsewhere, consistently find that regular unconditional payments result in greater spending on basic needs, modest reductions in work hours, and — contrary to the common assumption — continued labor force participation among most recipients. People receiving guaranteed income do not, on average, stop working.
The case for it: It is administratively simple. It does not require the government to predict which industries will be disrupted or which retraining programs will produce employable skills. It treats workers as capable of making their own decisions about how to navigate economic transition. It provides a floor that prevents displacement from becoming destitution.
The case against it: The cost is substantial. A universal payment of $1,000 per month to all American adults would cost approximately $3 trillion annually — roughly 75 percent of current federal discretionary and mandatory spending combined. Even a more targeted program for displaced workers only would require sustained funding at a scale that has no current legislative path. Critics from both the left and right have pointed out that the Musk-Altman version places the full fiscal burden on taxpayers broadly — including workers who have already been displaced and are paying into a system that their displacement has strained.
The deeper structural critique is that government-funded UBI socializes the cost of a disruption whose benefits accrue primarily to the companies and shareholders deploying AI. If the productivity gains from AI flow to corporations and their investors, and the safety net costs flow to the public treasury, the distribution of gains and losses runs in precisely opposite directions.
Position 2: An Automation Tax (The Robot Tax Proposal)
The automation tax — sometimes called a robot tax — attempts to close that gap directly. Rather than funding a safety net through general taxation, it levies the companies deploying AI to replace workers, directing the revenue toward displaced workers or retraining programs.
Bill Gates first proposed the concept in 2017. Senate Democrats formalized it in a 2025 report called The Big Tech Oligarchs’ War Against Workers, which projected that AI and automation could eliminate nearly 100 million U.S. jobs within a decade and proposed taxing companies that use AI to expand automation, with revenue directed to harmed workers. OpenAI’s own April 2026 policy blueprint — a 13-page document released the same month nearly half of all Q1 tech layoffs were attributed to AI industry-wide — included a version of the robot tax alongside proposals for a four-day workweek at full pay and a public wealth fund.
Tax scholars at Texas A&M have argued the levy is not just economically sound but structurally inevitable: the U.S. tax system is built on taxing labor. If labor is replaced by machines that pay no payroll taxes, the revenue base collapses. A robot tax is, in this framing, less a penalty on innovation than a correction for a structural gap that automation creates in the public finances.
A March 2026 academic paper from researchers at Penn and elsewhere found that of six policy responses evaluated — including UBI, upskilling programs, capital income taxes, and worker equity participation — only a Pigouvian automation tax set equal to the uninternalized demand loss per task fully corrects the market distortion that automation creates. Their argument: when companies automate en masse, they collectively destroy the consumer demand that all companies depend on, creating a negative externality that the market does not price. A tax internalizes that externality.
The case for it: It places the cost burden on the party generating the displacement and the profit. It preserves the tax base. It creates a direct financial link between the disruption and the remedy.
The case against it: Critics argue it disincentivizes innovation and reduces U.S. competitiveness relative to countries that do not impose similar levies. As one libertarian analysis put it, there was no Model T tax, no levy on the cotton gin, and no fine for using computers — and U.S. prosperity was built on embracing those transitions, not taxing them. Overseas competitors adopting AI without such levies could undercut American companies, ultimately costing more jobs than the tax saves. The definitional challenge is also real: what precisely counts as an automating displacement, and how is it measured and verified?
Position 3: Worker Equity Participation
A third model shifts from transfers to ownership. Rather than paying displaced workers through taxes or government checks, it proposes giving workers equity stakes in the AI systems that replace them — making workers shareholders in the productivity gains their displacement generates.
Sam Altman has proposed a version of this through his Worldcoin project, which aims to distribute a global digital currency. Vinod Khosla, a prominent Silicon Valley investor, has advocated for restructuring the tax code so that workers receive a share of AI-generated productivity directly. The concept draws on the logic of sovereign wealth funds — Norway’s oil fund being the most cited example — applied to AI productivity rather than natural resource extraction.
The case for it: It aligns incentives rather than simply redistributing income. Workers who own stakes in AI productivity have a direct interest in its success rather than a dependent relationship with the companies deploying it. It avoids the fiscal burden of government-funded UBI while providing workers with assets rather than transfers.
The case against it: Implementation requires either mandatory corporate restructuring or voluntary participation that history suggests will not materialize at scale. It also does not address the immediate income gap for workers displaced today, whose equity stakes — if they received them — would provide no near-term income security. The distribution mechanism remains largely theoretical.
Position 4: Retraining and Job Creation (The Incumbent Congressional Response)
The bills currently moving through Congress — the AI Workforce PREPARE Act, the Workforce of the Future Act, the AI Workforce Training Act — share a common framework: invest in retraining programs and tax credits for companies that offer AI career development, preparing workers for the jobs that AI creates rather than compensating them for the jobs it eliminates.
The Workforce of the Future Act authorizes $90 million in grants for workforce training. The AI Workforce Training Act offers tax credits to companies providing AI career development programs. The AI Workforce PREPARE Act directs the Department of Labor to study what a rapid adjustment assistance program for AI-displaced workers would look like.
The case for it: It addresses the skills mismatch directly. It preserves the labor force participation model rather than substituting transfers for work. It has more legislative momentum than any UBI or automation tax proposal currently before Congress.
The case against it: Ninety million dollars in training grants is not a serious response to an estimated $1.2 trillion in projected wage losses. The trade adjustment assistance programs that served as the model for these proposals — designed for workers displaced by trade agreements — have a documented record of limited effectiveness: participants in those programs historically had worse long-term employment outcomes than comparable workers who did not participate. Retraining assumes there are jobs to retrain into. If AI eliminates categories of work across multiple sectors simultaneously, the destination of the retraining is unclear.
What the Evidence Actually Shows
The honest summary of the research is that no single proposal has been tested at the scale the displacement problem requires, and the evidence that exists is partial.
UBI pilots show that unconditional payments do not destroy work incentives. Finland ran the world’s first nationwide, statutory basic income experiment from 2017 to 2018 — 2,000 randomly selected unemployed citizens receiving €560 monthly — and found modest increases in employment, significant improvements in wellbeing, and no meaningful reduction in labor force participation. What these experiments have not tested is fiscal sustainability at full population scale, because no country has yet implemented universal income covering its entire population rather than a targeted subset.
The automation tax has theoretical support and structural logic, but no country has implemented one at the scale proposed, and the competitiveness objection has not been resolved.
Worker equity participation is largely conceptual. The retraining programs with actual track records have underperformed.
What the evidence does show clearly is that the cost of doing nothing is also real and also distributed — to the workers who absorb the displacement, to the public systems that support them, and to the consumer demand that a generation of underemployed workers cannot generate.
The Question the Debate Has Not Answered
The policy proposals currently in circulation share a common feature: they differ primarily on who bears the cost of a transition whose benefits are already being captured by a specific and identifiable set of companies and shareholders.
Government-funded UBI places the cost on all taxpayers. Retraining grants place it on discretionary federal spending. An automation tax places it on the companies generating the displacement. Worker equity participation attempts to restructure ownership of the gains themselves.
Each of these is a legitimate policy position with documented arguments for and against it. What is not a position supported by evidence is that no response is necessary — that the labor market will self-correct at a pace and scale that makes the displaced workers whole without intervention.
The displacement is documented. The policy debate is live. The question of who should pay for it is the one that Congress has not yet answered — and the one that will determine, more than any specific program design, whether the transition to an AI economy produces broadly shared prosperity or concentrates its gains in the hands of the people who built the machine.
Amendment Media will keep tracking the votes.