ECB study says AI boom has not yet hit U.S. jobs broadly

Apocalyptic warnings about AI wiping out jobs are running ahead of the evidence. A European Central Bank study released on June 22 says the United States has so far avoided the broad labor-market shock many economists feared, with aggregate employment and wages still looking largely intact even as pressure builds in a few exposed corners.

The ECB paper, written by Isabella Moder and Til Pommer, used 2019 as its base year and tracked occupation-level employment growth through 2025. It found that jobs with a high risk of AI substitution declined by more than 4% on average over that period. The weakness showed up in occupations such as economists and graphic designers, but the central message was narrower than the doomsday forecasts that followed the rapid spread of generative AI: the technology is changing which workers feel strain first, not yet destroying employment across the board.

The bank said it focused on the United States because AI effects were likely to surface there earlier than in other major economies, thanks to early-adopting firms and a relatively flexible labor market. That makes the U.S. a useful stress test for the rest of the world. So far, though, the test has not produced a collapse in overall hiring or pay.

That view lines up with other recent research. Yale Budget Lab said it found no discernible broad labor-market disruption since ChatGPT’s release in November 2022. The St. Louis Fed said AI adoption is substantially higher in the U.S. than in Europe, but it did not find evidence that adoption is associated with job losses at the industry level. Goldman Sachs has said the effects are already being felt in niches of the economy such as tech, knowledge and creative work, yet no significant AI-led shift in the employment mix across the whole U.S. economy has appeared in the labor data.

The sharper warning is at the bottom of the labor market. The ECB said there is growing evidence that AI is negatively affecting employment for specific occupational sub-groups, particularly junior workers in highly exposed occupations. That fits a pattern in which entry-level tasks are automated, reallocated or compressed faster than senior jobs disappear. The bank’s own consumer survey data have also shown that younger workers and university-educated workers are more likely to use AI at work, suggesting the labor market is already splitting between workers who can ride the technology and those who are most exposed to it.

For now, the data point to rearrangement rather than wholesale replacement. A real AI labor shock would show up if the 4% decline in high-risk jobs spread into broader occupation groups, if junior hiring weakened more sharply, and if wage pressure moved from isolated pockets into the wider U.S. economy.