California AI-Unemployment Tracker (CAIT) - California Policy Lab

Skip to content

HomePublicationsLaborCalifornia AI-Unemployment Tracker (CAIT)

June 25, 2026

California AI-Unemployment Tracker (CAIT)

Labor

Ben Hyman, Till von Wachter, Diana Herrera, Roozbeh Moghadam, Jacob Morris, Swapnil Motghare and Kara Segal

Report: Tracking AI-Related Job Loss

Technical Appendix: Tracking AI-Related Job Loss

Press release: New AI-Unemployment Tracker

Table of contentsKey Findings from June 2026 Release<br>About the California AI Unemployment Tracker (CAIT)<br>Frequently Asked Questions<br>Contact Information

Share this research

&times;

Tracking AI-Related Job Loss Using Unemployment Insurance Claims Data in California

The California Employment Development Department partnered with the nonpartisan California Policy Lab at the University of California to conduct research to measure AI-related job loss trends in California, using Unemployment Insurance claims data, combined with AI exposure measures. The California Policy Lab developed the California AI-Unemployment Tracker (CAIT), which will be updated monthly. The underlying data are available for public use (see download link under the tracker). See the associated report for more information: “Tracking AI-Related Job Loss Using Unemployment Insurance Claims Data in California”

Download the data in Excel (ADA compliant)

Key Findings from June 2026 Release

Finding 1: No Evidence of Rising Statewide UI Claims in AI-Exposed Occupations +

Since the release of ChatGPT-3.5 in 2022, statewide UI claims through May 2026 show no evidence of a surge in AI-related layoffs. The proportion of UI claims coming from AI-exposed workers has also not seen a statistically significant increase relative to prior to the pandemic. This is consistent with existing Current Population Survey estimates that show no nationwide link between AI exposure and unemployment.

Finding 2: UI Claims Increased for College-Educated Workers under AI Exposure +

While AI-related job loss is not detected across all UI claims statewide, claims from college-educated workers in occupations with high AI exposure increased after ChatGPT-3.5’s release in 2022, and remained elevated through May 2026, compared to low-AI-exposure workers who did not experience a noticeable change in trend, relative to their pre-pandemic level in 2019 and early 2020.

Finding 3: UI Claims Rose in the SF Bay Area and Technology Sectors +

Over this same period, claims from workers in occupations with high AI exposure in the San Francisco Bay Area also experienced a sharp and sustained increase. AI-exposed UI claims were also particularly elevated statewide in technology sectors such as Information and Professional Services over the same period. Claims levels from many of these subgroups remained elevated in 2025 and through May 2026, even relative to their pre-pandemic level in 2019 and early 2020.

About the California AI-Unemployment Tracker (CAIT)

More information on data, definitions, methodology, and research partnership +

The California AI-Unemployment Tracker (‘CAIT’ or "The Tracker’) provides early signs of possible AI-induced job loss. It shows the number and share of Unemployment Insurance claims over time in California, categorized by how exposed to AI the person’s most recent job was (AI Exposure measure). For each unemployment insurance customer (also called claimants), the tracker additionally shows trends by age, education, gender, industry, race and ethnicity, and region, as well as the average AI exposure score by each group.

The tracker was created through a partnership with the California Policy Lab (CPL), a nonpartisan research center at the University of California. The data is updated monthly and analyzed to identify key findings. Its underlying tabulated data is available for public use. See below for definitions of terms used in the dashboard, its methodology, data source, and more.

Definition of terms

Initial Claims<br>The count of initial Unemployment Insurance claims filed by customers for each calendar month. Initial claims include both new claims (those applying for the first time) and additional claims, which occur when a customer returns to work, skips at least one certification week, and then reopens a prior claim before the benefit year expires. Each claim is dated to the month of its week-ending date.

Categorical AI Exposure Measures<br>The tracker uses two measures of AI exposure – Potential AI Exposure and Observed AI Exposure. Potential Exposure assesses whether large language models are capable of reducing the time required to complete an occupation’s tasks by at least 50%. Observed AI Exposure captures the extent to which an occupation’s tasks are performed using Anthropic’s Claude platform, based on observed Claude usage. More details about these measures are below.

Mean AI Exposure Score: The mean AI exposure score reflects how exposed to AI, on average, the customer’s job was.

Age:...