Data current as of August 11, 2026. Numbers on this page snapshot the dataset on the date above; the state pages refresh continuously as new WARN filings and news classifications land.
What "AI layoffs" means on JobShift
An AI-attributed layoff on JobShift is a WARN filing where two independent conditions are met: contemporaneous press coverage of the same company, and causal language in that coverage clearing a calibrated confidence threshold under the AI classifier. Filings without corroborating press coverage remain in the dataset but never carry an AI label. The full methodology is documented in the classifier deep-dive; the methodology page surfaces the live rejection and inclusion logic.
The consequence for the state-level numbers below: they measure provably AI-attributed layoffs, not the full universe of layoffs where AI might have been a factor. When a small employer lays off workers with no press coverage, that layoff is captured through WARN and appears in the state's total layoff count, but never in the AI-attributed subset.
2026 dataset at scale
Year to date through August 11, 2026, the JobShift dataset contains 2,885 US layoff events spanning 47 US states, the District of Columbia, and federal filings. Of those, 157 are AI-attributed — approximately 5.4 percent of the 2026 layoff volume the classifier has admitted to the AI category so far.
The historical footprint on layoff data goes back to 1988 for the earliest WARN filings, with over 64,000 layoff events on file in total. AI attribution is applied where the corroborating press-coverage window has sufficient density to make the classifier meaningful; the classifier itself is applied identically to every year in scope.
Top states by AI-attributed layoff events, 2026
The 2026 AI-attributed events cluster heavily in a small number of states. The table below lists every US state with at least five AI-attributed layoff events in the year-to-date window.
| State | AI-attributed events | Estimated workers | |---|---:|---:| | California | 89 | ~11,800 | | New York | 10 | ~830 | | Washington | 10 | ~5,800 | | Illinois | 9 | ~1,400 | | Maryland | 5 | ~740 |
Nine additional states — Arizona, Colorado, Delaware, Missouri, Nevada, New Jersey, North Carolina, Pennsylvania, and Virginia — each carry one or two AI-attributed layoff events for 2026. Three further AI-attributed layoffs are logged against federal or multi-state jurisdictions and do not roll up to any single state. An additional 20 AI-attributed events (roughly 13 percent of the 2026 total) lack a state-level region code in the current dataset and are excluded from the geographic ranking; they appear in the national count and on the AI layoffs view but not on any per-state page. Backfilling state attribution for those events is an open data-quality item.
The California concentration is severe: 89 of the 157 AI-attributed 2026 events (57 percent) sit in California alone. The next four states combined (New York, Washington, Illinois, Maryland) account for another 34 events (22 percent). The remaining 21 percent splits between the nine smaller states named above, federal filings, and the unattributed subset.
What drives the California concentration
Two structural factors explain most of it.
Silicon Valley continues to house a disproportionate share of the AI-adopting employer base. When the largest AI-adopting employers restructure, California-sited jobs are what those restructurings cut. The California state page breaks the 2026 events out by company and filing date.
The press-coverage requirement in the JobShift methodology skews AI attribution toward employers who get covered in tech press. That coverage is heavily West-Coast-tilted for structural reasons unrelated to JobShift. States whose 2026 layoffs are dominated by manufacturing, retail, or health-care employers appear in the total layoff count but not in the AI-attributed subset — not because those layoffs are not AI-driven, but because the corroborating coverage that would clear the classifier threshold does not exist. This limitation is described in more detail on the methodology page.
Year-over-year context
The comparison of partial 2026 against full 2025:
| Year | Total US layoff events | AI-attributed | AI share | |---|---:|---:|---:| | 2025 (full year) | 4,942 | 90 | ~1.8% | | 2026 (YTD through Aug 11) | 2,885 | 157 | ~5.4% |
The AI-attributed share of layoff events has roughly tripled year-over-year. Total US layoff volume, when the 2026 partial year is annualized to a full-year run-rate (approximately 4,700 events), tracks close to the 2025 total, so the shift is a composition change rather than a change in overall layoff volume. The direction is consistent with press-coverage patterns and matches the monthly trend.
What this data does not show
Cause versus timing. A layoff coinciding with AI adoption at the same company is not proof AI caused the layoff. The classifier requires causal language in the corroborating press coverage to avoid conflating timing with causation. When the language is available and clears the threshold, the label is applied; when it is not, the layoff is left uncategorized.
Undercounted small employers. Filings under the federal WARN threshold (100 workers or 50 at a single site) do not appear in the WARN feed at all, and mid-size employers whose news coverage does not reach the tech press cannot receive AI attribution. Both effects push the AI-attributed count below the true count.
Attribution lag. A layoff in the current month may not receive its AI label until press coverage catches up. Recent-month counts are always undercounts and settle upward for four to six weeks after the filing date.
Live per-state data
The state-level counts on this page snapshot the dataset on August 11, 2026. For real-time per-state numbers as new WARN filings and news classifications land:
- The states index links to all 50 state pages.
- The California page covers the state with the highest 2026 AI concentration.
- The AI layoffs view shows the monthly time series with company-level detail.