Most Californians believe artificial intelligence will reshape their work, their communities and the nation’s politics, but anxiety about the technology runs deep, and few expect it to create more jobs than it takes, according to a Carnegie Endowment for International Peace survey released last month.

The 2026 Carnegie California AI Survey, now in its second year, polled 1,500 adult residents through YouGov from April 21 to May 11, in English and Spanish. The results carry a margin of error of plus or minus 3.5 percentage points.

Use of the technology at work is spreading fast. About six in ten Californians said they use A.I. daily or occasionally on the job, up from a year earlier, and the share who said they rarely or never use it fell by 15 points. About half described A.I. use at their workplace as extensive or moderate, also an increase. The payoff, so far, is modest: most said the tools brought minor or no improvement in productivity.

The bigger worry is jobs. Californians are increasingly concerned about losing work to A.I., and few expect the technology to open up more opportunities than it closes. The concern comes with an unusual bipartisan overlap: a majority supports redistributing profits from large companies into workforce training and A.I. skills programs, a similar share said it was eager for such training, and Republicans and Democrats alike back a state policy protecting workers from displacement. About half expect to use A.I. tools more at work in the future.

Model safety, the survey found, is no longer a niche concern. Nearly three-quarters of Californians agreed that the government should require A.I. companies to test their most advanced systems for safety and provide a detailed plan for preventing harm.

Asked about specific risks, residents ranked privacy first, followed by cybersecurity, spam or fraud, misinformation and deepfakes. Cybersecurity climbed from fourth place in 2025 to second this year, even though the share citing it as a concern held steady.

Residents are far less sure about A.I. in government hands. Under 10 percent supported federal, state or local agencies using the technology to make decisions that affect them, though they were more open to it in cases where such use definitively improves decision-making.

Confidence in spotting machine-made content is growing. About half of Californians said they were at least somewhat confident they could tell real material from A.I.-generated material, up notably from last year. Concern about election deepfakes and A.I.-fed polarization remains widespread, though it has eased somewhat since 2025.

The survey found divides by age, education and income — workplace use varied with age and schooling but was similar across genders and party lines — alongside broad agreement between the parties that A.I. matters to the nation’s and the state’s economic competitiveness. Most respondents said large businesses would benefit, and a growing share said small businesses would, too. The findings arrive as candidates, policymakers and scholars float bigger changes to the social safety net, from centralized wealth pools funded by private companies and taxes on top-earning companies to government job-loss trackers and universal basic income.