Rapid AI could cut U.S. labor force by 10 million

Researchers at the Federal Reserve Bank of Chicago, the Forecasting Research Institute, Yale, Stanford and the University of Pennsylvania found that rapid progress in artificial intelligence could lower U.S. labor force participation to 54% by 2050, with roughly 10 million jobs lost directly to AI rather than demographic change.

The paper is based on a survey of 69 economists, 52 AI specialists and 38 superforecasters. Respondents across the groups agreed that faster AI development would reduce labor force participation. The report described that outcome plainly as “fewer people working.”

The authors defined the report’s “rapid” scenario as one in which AI systems outperform humans on most cognitive and physical tasks by about 2030. Examples cited in the paper include systems that can negotiate book contracts, provide broad factory and home assistance, and take over roles such as freelance software engineers, paralegals and customer service agents.

Under the rapid scenario, economists in the survey projected U.S. annual GDP growth near 3.5% by 2045–2049. AI specialists were more optimistic, projecting about 5.3% growth over the same period. The paper also projects a concentration of wealth: the top 10% of households could own roughly 80% of total wealth by 2050 in the rapid scenario.

On policy responses, surveyed economists showed strong support for targeted retraining programs, with 71.8% in favor. Fewer economists supported broad measures: 13.7% backed job guarantees and 37.4% supported universal basic income. The report notes that public opinion is more open to structural interventions than the economists surveyed.

The authors reviewed recent labor data and other studies and reported mixed short-term signals. A late-2025 study cited in the paper found no clear sign of mass unemployment nearly three years after large language models became widely available. Other work cited shows early impacts for specific groups: employment among workers aged 22 to 25 in the most AI-exposed occupations fell about 13% relative to less-exposed peers.

The paper highlights that expert disagreement centers less on whether powerful AI will arrive and more on how the economy will look when it does. Economists remain divided on whether AI will generate new categories of work or automate the creation of new tasks.

The authors say optimal policy choices depend on which scenario materializes and emphasize uncertainty about the timing, scale and distributional effects of AI-driven change.