Vice President JD Vance told the Daily Wire’s Michael Knowles recently that American conservative economic policy is now much more Alexander Hamilton than Milton Friedman, that the free market policies of the 1980s were fine for their era but the party has moved on, and that President Trump has already completely reoriented the economic conversation toward what Vance called an American developmentalist approach. He noted approvingly that when Trump suggested seizing equity stakes in AI companies, nobody really even protested. OpenAI CEO Sam Altman appears to agree, filing for an IPO that would reserve five percent of shares for the federal government or a sovereign wealth fund and issuing a thirteen-page white paper calling for expanded government control of AI, higher taxes on capital gains and corporate income, taxes on automated labor, and government incentives for a thirty-two-hour work week with no loss in pay.
Liya Palagashvili, senior research fellow and director of the Labor Policy Project at George Mason University’s Mercatus Center, joined Dan Proft on Chicago’s Morning Answer to assess whether this convergence of left and right around industrial policy is warranted by the data and what AI is actually doing to the American workforce.
On the prospect of government taking equity stakes in AI companies, a practice China employs through what are known as golden shares, Palagashvili said it would vastly change the dynamics of the American economy and its institutions. She said the United States has been built on a fundamentally different institutional mix characterized by limited government that allowed freedom for workers and businesses to grow on their own, and that moving toward state ownership stakes in private companies would represent a departure from the principles that produced American economic dominance. She noted that major AI companies including OpenAI and Anthropic have recently begun walking back their previous predictions about AI’s catastrophic impact on jobs, quietly abandoning the doomsday narrative that half of all jobs would disappear.
On what is actually happening in the labor market, Palagashvili presented research findings that tell a far more nuanced and in many ways more optimistic story than either the catastrophists or the industrial policy advocates suggest. She cited a study of twenty-one thousand US businesses that found firms adopting AI heavily grew headcount by ten percent over the two years following adoption, while firms with low AI adoption saw no statistically significant change in employment. At least in the current phase, heavy AI adoption appears to expand the human workforce rather than replace it.
Her own research, however, points to what she considers the most fascinating and underappreciated dynamic: AI is making it significantly easier to work for yourself rather than inside a traditional firm. She examined Census Bureau monthly business application data, which distinguishes between applications likely to hire employees in the future and those unlikely to do so. In AI-exposed sectors, business applications planning to hire employees actually fell since 2024, but business applications with no plans to hire, the best real-time measure of solo business formation, rose twenty-seven percent. All of the net growth in business formation in AI-exposed industries over the past two years came from people starting businesses to work for themselves, not to build companies with employees.
She said the mechanism is straightforward. AI tools are lowering the cost of operating without a firm by enabling a single person to perform functions that previously required a team: research, drafting, design, coding, bookkeeping, pitch deck creation, data analysis. A consultant who once relied on junior analysts, editors, and designers can now produce comparable work independently. A researcher or economist can produce near-publication-ready work with far fewer institutional resources than was possible before AI. In industries where AI handles the cognitive support work, the result is not unemployment but a growing ability to work outside traditional corporate structures.
She said this pattern is showing up not only in business formation data but in monthly labor statistics as well. Solo self-employment in occupations most exposed to AI jumped twenty percent between 2022 and 2026, concentrated in professional services, information technology, and other AI-adjacent sectors. In occupations that AI barely touches, such as construction and wholesale trade, the growth in solo self-employment barely moved.
On the longer-term question of whether AI will eventually crowd out the benefits it is currently producing as unsupervised machine learning becomes more sophisticated, Palagashvili said the possibility is real for some specific jobs and tasks. But she said the history of technological shocks demonstrates that new industries and roles emerge that cannot be foreseen at the moment of disruption. She cited the example that in 1900 there were no jobs for airplane pilots, yet within a century an entire industry employing tens of thousands had been created from a technology that did not exist. She said the same pattern has repeated with every major technological transition, and that the critical metric is net job growth rather than the disappearance of any individual occupation. She said the evidence so far points toward net growth rather than net destruction, which is precisely why the AI companies themselves are quietly abandoning the apocalyptic workforce predictions they were making as recently as two years ago.