AI Job-Loss Fear Doubled—and Worried Workers Expected to Save Less

The share of surveyed U.S. workers worried that artificial intelligence could cost them their own job rose from 5% in December 2024 to just over 10% in December 2025. On September 2, 2026, the Federal Reserve Bank of Boston released two Current Policy Perspectives briefs linking that increase in concern with weaker expectations about saving, particularly among workers already struggling to afford the same goods and services.

An independent report published September 3 also documented the doubling of personal job-loss concern. But the findings measure perceptions and plans—not verified displacement, account balances or actual saving—and do not establish that fear of AI caused financial strain.

Four signals define the survey’s risk-perception matrix

The results distinguish four questions that can otherwise blur together: whether a worker fears losing their own job, whether they expect disruption across their industry, whether AI appears to improve their productivity and whether they expect their household saving rate to fall. Each signal describes a different form of exposure.

• Personal job risk: The narrowest measure captures workers who agreed that they worried about losing their own position because of AI.
• Industry disruption: A worker could feel personally secure while still expecting AI-related layoffs or lower employment elsewhere in the industry.
• Productivity and security: Concern was greatest among AI-exposed workers who did not perceive either a clear productivity benefit or complete distance from the technology.
• Household finances: Expected saving was weakest when job-loss concern coincided with reduced purchasing power.

The primary worker-perspectives brief shows why these categories should not be collapsed. In the 2025 wave, 10% feared losing their own job, while 60% expected AI-related layoffs or fewer workers in their industry; personal concern reached 23% in consumer services, 21% in leisure services and 15% in firm services. Among employed respondents with some AI exposure, the estimated likelihood of job-loss concern was 21.2% for those neutral about AI improving their productivity, compared with 6.4% among those who strongly disagreed and 6.1% among those who strongly agreed.

The industry percentages are shares expressing concern, not forecasts of how many jobs will disappear. Likewise, the personal measure records a belief about possible loss; it does not show that the respondent’s employer had announced or completed a displacement.

The productivity middle felt the least secure

The relationship between perceived productivity and security was hump-shaped rather than steadily worsening as workers encountered more AI. Workers at both ends of the productivity scale expressed relatively little concern, while anxiety peaked among those who were neither clearly benefiting nor clearly unaffected.

One possible explanation is that workers in the middle had started allowing AI to substitute for parts of their work without seeing an obvious improvement in what they produced. By contrast, people reporting strong gains may have viewed the technology as increasing their value, while those reporting no gains may have worked in roles where AI had little immediate relevance.

That interpretation remains a hypothesis, not an observed mechanism. Productivity, task substitution and job-security concern were self-reported within the same survey framework, so the results cannot determine which perception came first. Workplace conditions omitted from the survey could also influence all three responses.

Affordability pressure accompanied weaker saving plans

The companion saving brief defines the outcome as whether workers expected to save a smaller share of earnings over the following 12 months. That share rose from 11% in the 2024 wave to 21% in 2025; after the available demographic, financial, geographic and employment characteristics were considered, workers worried about AI-related job loss were roughly 8 percentage points more likely to expect a lower saving rate. Among workers reporting reduced purchasing power, 35% expected a lower rate, compared with 11% of those able to afford the same goods and services, and respondents facing both affordability difficulty and AI concern were nearly twice as likely to expect lower saving as affordability-constrained respondents without the AI concern.

This pattern runs in the opposite direction from a simple precautionary-saving story, in which fear of income loss prompts a household to put more money aside. A worker already under budget pressure may lack room to increase saving, regardless of how strongly that worker wants a larger financial buffer.

The association is therefore compatible with several explanations. Existing financial strain could make both employment and future saving feel more precarious; expectations about income could shape both answers; or other circumstances absent from the model could influence them together. Statistical controls address some observable differences but do not turn the relationship into evidence that anxiety caused distress.

The distinction in the headline is consequential: worried workers expected to save less. The survey did not follow their bank accounts over the subsequent year, calculate realized saving rates or determine whether their plans became behavior.

The survey tracks sentiment, not an AI displacement count

The findings come from employed respondents answering special modules of the New York Fed’s nationally representative, internet-based Survey of Consumer Expectations. The survey uses a rotating panel of approximately 1,300 U.S. household heads, and the relevant modules were fielded in December 2024 and December 2025.

That design provides two snapshots of workers’ views, but it does not by itself prove that the same individuals became more fearful between waves. Estimates for smaller industry, education or demographic groups also carry greater uncertainty, particularly where only a limited number of respondents fit the category.

Separate evidence reinforces the need to keep perceived risk apart from realized labor-market outcomes. An August 2026 Stanford Institute for Economic Policy Research paper found substantial worker concern in a different multi-wave survey but no statistically significant response in job postings or layoffs among more exposed occupations during the period studied.

As of the September 2 release, the Boston findings establish that personal AI job-loss concern increased and that worried workers—especially those under affordability pressure—had a weaker saving outlook. Whether those expectations translate into displacement, reduced income or lower realized household saving will require later evidence connecting survey responses with subsequent outcomes.

Also read:

• Top 5 Tips Gig Economy Workers Can Save for Retirement
• Quasa Connect: The Optimal Choice for Hungarian Remote Workers to Earn in Crypto
• The Great Switch: LinkedIn Emerges as the World's Top Dating Network, While Dating Apps Turn into Job Hunt Hotspots