AI Job Loss Dashboard: What the Wage Data Shows
Explore what the AI job loss dashboard estimates for US wages, how its state figures are calculated, and what the data can—and can’t—tell you.
AI Job Cuts, Measured in Dollars: What the State Dashboard Shows
Layoff totals are easy to repeat and hard to interpret. A company announces job cuts, says AI played a part, and suddenly the number turns into another talking point in the automation debate. But a headline figure doesn't tell you what the cuts might mean for workers, households, or the businesses that rely on that local spending.
The Center for AI Safety's Wage Loss Dashboard tries to make that impact more concrete. It pairs AI-attributed layoff announcements with state wage and population data to estimate annual wage losses across the United States. In its latest snapshot (data through September 2026), it estimates just over 117,000 AI-attributed job cuts nationwide and about $8.5 billion in annual wages tied to those cuts.
Those numbers are worth looking at - but they're estimates. They're not a verified count of every person replaced by AI. That distinction matters. Below is what the dashboard measures, how the state estimates are built, where the approach holds up, and where readers should be cautious.
What the AI job loss dashboard shows
The Wage Loss Dashboard shows AI-attributed job cuts by month and state, along with estimates of the annual wages associated with those cuts. You can view the national totals or focus on a specific state. The point is to translate layoff announcements into an economic measure that's easier to connect to household finances and local economies.
In aggregate, the dashboard's latest figures are large: just over 117,000 US job cuts attributed to AI in 2026, and an estimated $8.5 billion in annual wages. It also calculates an average annual wage loss per resident. Nationally, that comes out to about $25 per person. California is higher (about $56), and Alaska is lower (about $14).
It's easy to misunderstand these per-person numbers. They don't mean every Californian lost $56 in income, or that any single worker's pay dropped by that exact amount. The dashboard takes an estimated state-level wage loss and divides it by the state's population. The result is mainly for comparing estimated impact across places, not for describing what happened to each resident.
Why state-level estimates differ
The dashboard attributes just over 27,000 job cuts to AI in California, versus 138 in Alaska. States with big technology and media sectors may see more announcements in those areas, and average wages differ widely across states. The cut counts and the wage assumptions both affect the final estimate.
That's why the state comparisons work best as a broad indicator of where the estimated exposure is concentrated. They're not a precise ranking of where AI-driven unemployment has hit hardest. The dashboard also makes an important claim: its state estimates aren't "observed" AI layoffs. Instead, it takes a national monthly share of cuts that cite AI and applies it to each state's announced job cuts.
Why translate layoffs into wages?
A layoff count tells you how many positions an announcement claims are affected. A wage estimate adds a rough financial dimension by expressing the annual earnings associated with those positions. For readers, that can shift the conversation from "how many jobs" to "how much income is no longer coming in," and what that could mean for local communities.
Still, it's only part of the story. Wages tied to eliminated positions aren't automatically equal to lost household income for a full year. Some people land new jobs quickly, receive severance, shift into different roles, or leave the workforce. The dashboard doesn't follow outcomes for individual workers. Its value is in making a potential cost visible - not in providing a complete picture of what happens next.
How the wage loss estimates are calculated
The dashboard combines multiple data sources. Challenger, Gray & Christmas track announced job cuts across the United States, including state-level figures and national totals that companies publicly link to AI. Wage inputs come from the Bureau of Labor Statistics' annual mean wage data for each state. Population estimates come from the US Census Bureau, which the dashboard uses to calculate the average loss per resident.
At a high level, the method looks like this:
- Collect announced job cuts by month and state.
- Determine the national share of cuts that companies have publicly attributed to AI.
- Apply that monthly share to each state's announced cuts to estimate AI-attributed cuts there.
- Multiply estimated cuts by the state's average annual wage to estimate annual wage loss.
- Divide the state total by its population to get an average per-person figure.
This is a straightforward way to turn incomplete layoff reporting into comparable numbers. It's also why the output should be described as estimated wage loss, not as a confirmed count of wages that disappeared because of AI.
The attribution problem
There's no single standard definition for what makes a layoff "AI-related." Companies might cite automation, productivity, restructuring, or a broader shift in how work gets done. Sometimes executives explicitly connect the cuts to AI; other times the link is more indirect. Either way, the company's explanation helps show how it frames the decision, but it doesn't independently prove that AI made a specific role unnecessary.
That's because the dashboard relies on publicly reported company explanations to support its estimates. It doesn't require proof that a particular employee was replaced by a particular AI system. The intent is to track a broader pattern of cuts companies associate with AI, while trying to exclude layoffs that appear unrelated. It's a pragmatic approach - but it can't eliminate ambiguity in the underlying announcements.
Jobloss.ai, a separate project maintained by the Alliance for Secure AI, uses different inclusion criteria. It logs qualifying large layoffs when there's some mention of AI or related terms like automation or productivity. Different trackers can therefore produce different totals without any obvious "math error." They may be using different definitions, thresholds, or source material.
What the wage calculation leaves out
Using average annual wages helps keep the estimate consistent, but the reality is messier: workers affected by layoffs don't all earn the state mean wage. Pay can be higher or lower, and the data doesn't capture differences in role, seniority, hours, or benefits. The calculation also doesn't show how long someone remains unemployed or what they earn in a new job.
So the dashboard is answering a narrower question: What annual wage value might be associated with the AI-attributed cuts, based on the stated assumptions? It isn't answering how much income workers will actually lose over time. Keeping those questions separate makes the data easier to use responsibly.
What the numbers can - and can't - tell you
The dashboard helps because it turns a fast-moving, unevenly reported trend into something easier to compare. It can show the scale of announced cuts, where estimates are concentrated, and how changing wage assumptions might alter the discussion. It also gives workers, local leaders, and policymakers a shared starting point - something that doesn't rely only on anecdotes.
But it isn't a full model of AI's effect on employment or the broader economy. CAIS says the dashboard doesn't include jobs created by AI, productivity gains, re-employment, or secondary economic effects. Those gaps matter. If a company reduces headcount while increasing output, the dashboard may capture the attributed cut without capturing the productivity gains. And if an affected worker finds a comparable job quickly, the estimated annual wage value may overstate eventual income loss.
There's another risk in the other direction, too. Announced layoffs won't capture every change AI brings to work. A business might slow hiring for certain roles, shift more tasks to contractors, or reorganize without issuing a large layoff announcement explicitly tied to AI. In that sense, the dashboard is a partial view in both directions.
Announced cuts are not proof of direct replacement
A company can cite AI as one factor in restructuring without showing that an AI system took over the exact work of every eliminated role. Layoffs can also reflect changing demand, cost-cutting efforts, mergers, or earlier over-hiring. In real organizations, multiple factors often point toward the same decision.
That doesn't mean the AI link is meaningless. Public statements create a record of how companies explain their decisions, and repeated references to AI are worth tracking. The more careful interpretation is that these figures measure cuts publicly associated with AI, not a scientifically isolated count of jobs directly replaced by AI.
The economy-wide effect is still unsettled
There's no consensus on whether AI will create enough new work and growth to offset displacement - or how quickly those effects might show up. The answer likely varies by industry, occupation, region, and time horizon. An efficiency gain in one part of an economy can reduce demand for certain tasks while increasing demand elsewhere - or it may simply let the business produce more with the same team.
That's why a layoffs dashboard shouldn't be treated as a net employment forecast. It captures one side of a transition. By itself, it doesn't show how displaced work compares with new jobs, higher output, lower prices, or changes in wages.
What the dashboard means for workers and businesses
The most immediate point is that job cuts affect more than the people named in a press release. Wages pay for rent, groceries, transportation, childcare, and local services. If enough workers in a region lose income at the same time, some of that impact can spread to shops, service providers, and landlords. The effect depends on how quickly people find work and whether replacement roles pay at similar levels.
This is where the dashboard's dollars-and-cents framing can help, even with the limitations. It gives local stakeholders reasons to ask more specific questions: Which sectors are affected? Are displaced workers finding new jobs? Do replacement roles stay in the same region? What training or transition support exists? A national total can't answer those, but it can point to where closer local analysis is needed.
The separate signal from remote freelance tasks
The dashboard also includes information from the Remote Labor Index, a benchmark co-developed by CAIS. The index tests AI agents on real freelance tasks, including work like web development and video production. In the researchers' cited results, the best-performing agent completed 2.5% of projects at professional quality when the benchmark launched. Eight months later, that figure rose to 15.8%.
It's a relevant result, but it measures capability - not market adoption. A benchmark can show that an AI agent can complete a task to a specified standard. It doesn't show that clients have actually assigned that work to an agent, stopped hiring freelancers, or decided the output is reliable enough for production. Capability trends can hint where work might shift, but they don't prove automation is already happening across the board.
For businesses, the practical distinction is clear. Testing an AI agent in a contained workflow isn't the same as removing human review from a customer-facing process or something financially sensitive. Quality control, error handling, data security, and accountability all come with real costs. A task that looks automatable in a benchmark may still require human judgment in the real world - especially when requirements are incomplete or the cost of mistakes is high.
A practical way to read the dashboard
If you're using the figures for planning or reporting, keep your claims aligned with the method. A simple checklist is:
- State the time period. The cited national and state figures reflect the dashboard's latest reported snapshot, with data through September 2026.
- Call the figures estimates. State-level AI cuts are inferred using a national monthly share; they are not individually observed AI layoffs.
- Separate cuts from net employment. The dashboard does not count job creation, re-employment, or productivity gains.
- Treat wage values as an approximation. Average state wages don't represent each affected worker's salary or final income loss.
- Look for local evidence before making a local claim. State-level estimates can point to where further investigation is needed, but they don't explain what happened to a specific town, industry, or worker.
This avoids two common mistakes: dismissing the data just because it's imperfect, or presenting an estimate as if it were a definitive count. The middle ground is to treat it as a signal that leads to better questions.
Conclusion: use the figures as a signal, not a verdict
The Wage Loss Dashboard adds useful context to AI layoff reporting by translating announced cuts into estimated annual wages and state-level comparisons. Its main contribution isn't that it settles the debate over AI and employment. It helps make one part of that debate more visible - and it shows that the impact may be unevenly distributed.
These numbers are built from public announcements, wage averages, and population data. They can't prove that AI directly replaced every affected worker, calculate each person's eventual income loss, or predict net employment after new jobs and productivity gains are taken into account. Those limits are part of what the data can and can't do.
If you're tracking AI's effect on work, use the dashboard as a starting point. Verify the time period and assumptions, compare the estimate to local employment data, and look for evidence on what happens to workers after a layoff. That's generally more useful than treating any single national total as the final answer. Read the original ZDNET report for the source figures and dashboard context.