Wholestory

The Bureau of Labor Statistics published its biennial count of displaced workers — the first covering a full three years after ChatGPT. From January 2023 through December 2025, 7.4 million Americans lost jobs to plant closings, abolished positions or insufficient work, up from 6.3 million; of the 3.3 million long-tenured among them, 66.1 percent were back at work by January 2026, little changed. The survey offers three reasons for a job loss and none is technological. The growth was in durable-goods manufacturing, up 215,000, not the white-collar occupations AI is said to be hollowing out.

Uber said it would eliminate about 3,300 roles, roughly 10 percent of its global workforce, in its largest layoff since the pandemic. Chief executive Dara Khosrowshahi framed the cut as removing management layers, halving one- and two-person teams and shedding roles more than seven layers below him to make the company leaner and faster; remote work was cut to under 1 percent of staff. AI is not named as a reason. Coming the same week Challenger reported restructuring — not AI — back atop the stated reasons for August's layoffs, it is the month's marquee case of a large, named cut whose announcement reaches for every word except the one the headlines supply. It is separate from and larger than the customer-service cut Uber made in July.

U.S. employers announced 52,881 job cuts in August, the lowest August total since 2022 and down 38 percent on a year earlier, and for the first time since February artificial intelligence did not lead the stated reasons. Challenger, Gray & Christmas counted 3,462 cuts attributed to AI — its lowest monthly figure since December 2025 and the end of a five-month run, March through July, when AI topped the list every month. Restructuring returned to first with 16,173. The cumulative claimed line still rises, because it only ever adds: AI has now been cited in 116,175 announcements in 2026, about 22 percent of the year's cuts, and 188,000 since 2023, and it remains the leading reason year-to-date. But the monthly flow that built the line has fallen away, in the first full month whose largest cuts — Uber's among them — reached for other words. Announced hiring plans reached 119,825 year to date, up 37 percent, though the roles are not being filled quickly.

Three years of the New York Fed's regional business surveys, refreshed in August 2026, found AI adoption now widespread but the workforce response muted. 61 percent of service firms and 51 percent of manufacturers reported using AI, up from 40 and 26 percent a year earlier. Yet only 4 percent of service firms said they had laid anyone off because of AI — up from 1 percent — and no manufacturers reported AI layoffs in either year. About 15 percent of service firms hired fewer workers because of AI, while 13 percent hired more to help them use it, and retraining outnumbered replacement across the board. The economists' conclusion is the corroboration layer's steady refrain: AI 'has been more likely to augment workers than replace them.'

Dallas Fed economists linked millions of Texas job postings to a task-based measure of how far generative AI can automate each occupation. Since ChatGPT's late-2022 release, openings for more-exposed occupations fell about 5 percent relative to less-exposed ones by end-2023 and about 8 percent by early 2025. A panel of surviving incumbent firms cut the same postings 8 to 9 percent by early 2026, so new AI-native firms and company failures are not the cause. Across Texas, AI exposure is estimated to have trimmed total job postings by 1.8 percent in 2024 and 2.6 percent in 2025 — a hiring pullback rather than dismissals, landing hardest on recent graduates and job-switchers.

A Goldman Sachs research report (economists Sarah Dong and Joseph Briggs) finds AI starting to weigh on labor markets across the major developed economies. Industries most exposed to AI automation — information and communication services, call centres, software publishing, management consulting and advertising — have seen job-openings growth slow and employment drift below trend since the second half of 2022, most pronounced in Germany, Australia and the United States. Call-centre employment now sits 39 percent below trend in the US, 33 percent in Canada and 27 percent in Germany. The bank puts AI adoption across developed economies at roughly 15 to 20 percent, with junior workers facing the strongest hiring headwinds. Economy-wide, though, the effect stays limited: a 10 percent occupational exposure to AI is associated with only a 0.1-percentage-point drag on annual headcount growth. It extends the entry-level erosion this page tracked in US data into an international picture — a correlational read of exposure against openings.

Reuters obtained the internal planning for Project OT, Meta's "AI-native" reorganisation: agents doing more, teams doing less, some cut by as much as 60 percent through layoffs, freezes and performance exits. Meta made 10 percent cuts in May and then dropped the broader round, because the tools did not deliver. The company's own numbers say why. Code changes to Meta's internal platforms rose 220 percent year on year, per an internal post by CTO Andrew Bosworth — and features actually shipped to users rose 36 percent. Major technical and security incidents rose 40 percent, and time spent responding to them 70 percent. An April post warned that agents without sufficient oversight were taking "large-scale, disruptive actions that humans are unlikely to execute". Meta declined to comment on the figures.

Revelio Labs and the consultancy System2 took the companies that publicly blamed layoffs on AI — Meta, Amazon, General Motors, The Washington Post, UPS, McKinsey among them — and checked their hiring against size-matched peers. The cohort had grown AI headcount by a median of over 11 percent in the two prior years while cutting non-AI roles by more than 3 percent; peers grew AI roles faster, over 13 percent, but held non-AI headcount flat. AI roles were about double the peer share of the workforce before the announcement and still double six months after, so the shift was real and it stuck. Their hiring concentrated in IT infrastructure, product strategy, R&D and cybersecurity. And then the other finding: nearly half the cohort shrank overall or cut AI headcount before announcing AI layoffs, and now trails its own industry in AI adoption. Nobody observed a productivity gain.

A Washington Center for Equitable Growth brief says plainly what this page's structure assumes: US federal statistics cannot currently show whether or how AI is changing the labour market. Data on unemployment, wages and job availability is fragmented, often untimely and hard to link. Firm-level AI adoption is asked about in binary yes-or-no questions, inconsistently between surveys, and never connected to what happens to those firms' workers. The warning for what comes next: Connecticut's WARN-AI disclosures begin on 1 October, and a disclosure requirement not linked to employment outcomes produces more claims, not corroboration.

Three of the month's largest announced US technology cuts arrived with no AI attribution at all. Apple let go more than 200 people across its Vision Products Group, Siri and Intelligent Systems Experience teams, reported as the beginning of putting Vision Pro on ice and a re-architecture of Siri. TikTok will cut 250 and close its Nashville office in October, saying only that it wants "to streamline our operations and better align our teams for long-term growth" — no reason, no named team. Zillow will cut just over 500, its chief executive citing "a disciplined cost structure" and "getting more efficient". Challenger has had AI leading its stated reasons for job cuts five months running. These three chose other words, recorded verbatim so they can be checked against the filings later.

A bill that would create the most ambitious attempt yet to fill the missing column passed both chambers of the New York legislature in June and, as of 17 August, had still not been sent to Governor Kathy Hochul for signature. Assembly Bill A9581B, introduced on 21 January by Assembly Member Bronson, would add a new section to state labor law requiring every business with more than fifty employees doing business in New York, plus every publicly traded business, to file a report with the Department of Labor by 1 March each year estimating how many employees it displaced or cut hours for, how many it hired or gave more hours to, and how many previously filled positions it decided not to fill — in each case 'due in full or in part to use of artificial intelligence.' Employers would also describe what they use AI for, how they oversee it, how often and how long they use it, and whether it touches sensitive personal data. Failure to file carries a civil penalty of up to $500 a day, with ninety days to cure after notice, and the Department of Labor must publish an aggregate analysis by sector, geography and business size within 120 days of each deadline. Lawmakers widened the bill before passing it, dropping the coverage threshold from more than a hundred employees to more than fifty. What makes it different from the WARN amendments moving elsewhere is that it does not wait for a mass layoff. It reaches the quieter arithmetic — the vacancy left unfilled because the remaining team can absorb the work with AI — that no layoff notice ever captures. New York has already been running the narrower experiment, after Governor Hochul's 2025 State of the State directed the Department of Labor to ask employers filing WARN notices whether a layoff was AI-related. Its 2026 record so far contains exactly one action expressly citing artificial intelligence: a Nespresso notice covering 46 workers at a New York City location, which the department lists under the reasons 'Relocation of Business, Artificial Intelligence.' The public record does not say how much of it was relocation and how much was AI, or whether AI replaced anyone directly. That single ambiguous line is the whole difficulty of the phrase 'due in full or in part' — two firms with identical workforce changes could answer differently and both be telling the truth. California is pursuing the narrower route in parallel: SB 951 would require a Cal/WARN notice for a layoff caused 'in whole or in substantial part' by an automated system to specify which job functions are being automated, and add a separate technology-hiring-disruption notice to the Employment Development Department, with the same $500-a-day penalty; it remained in progress on 13 August. Connecticut's disclosure duty, the only one enacted, takes effect on 1 October.

Erik Brynjolfsson, Bharat Chandar and Ruyu Chen released a revised 'Canaries in the Coal Mine?', extending their ADP payroll panel — 3.5 to 5 million workers observed monthly — through June 2026. The two-sided result is the most careful account yet of what the payroll data can and cannot show. Aggregate employment in the sample grew about 6% since November 2022, against about 4% in the most AI-exposed fifth of occupations: no economy-wide displacement is visible. But employment of 22-to-25-year-olds in AI-exposed occupations now sits 19% below where it would be had it kept pace with their less-exposed peers, up from 15% on the same measure a year earlier, and there is no comparable gap for experienced workers. In levels, employment for that age group fell about 11% in the two most exposed quintiles while growing about 10% in the three least exposed. The adjustment runs through hiring, not firing — separations fell at least as much for exposed young workers as for everyone else, the opposite of what displacement would look like — and shows up in headcount rather than base pay. The revision adds mechanism. Splitting occupations by whether observed AI use automates a task or complements the worker, the automation coefficient for 22-to-25-year-olds is -0.098 per standard deviation and shrinks steadily with age, while the complementarity coefficient turns positive and significant for workers past 41. Employment has fallen among young workers in jobs built on codified knowledge — the formal, documented kind a textbook can teach — and risen among experienced workers in jobs built on the tacit knowledge that comes only from practice. The authors also publish the finding that most coverage omits, and it cuts against their own headline: the occupation-level estimate for the most exposed young workers is -0.179 with no controls, but attenuates to -0.091 once the college share of an occupation is controlled for, and to a statistically insignificant -0.080 with interest-rate, education and remote-work controls together. They argue education may be the channel AI operates through rather than a rival explanation, but they do not claim to have settled it, and they note the divergence is sharper in their payroll sample than in national survey benchmarks. One number to treat carefully: the headline metric changed between versions. Earlier releases led on regression estimates of 13% and then 16%; this one leads on a simpler descriptive divergence of 15% rising to 19%. The widely repeated 13-to-16-to-19 progression is not a like-for-like series.

A U.S. Census Bureau working paper by the economist Lee C. Tucker did what no official series had yet done — it took the government's own matched employer-employee administrative records and measured the AI-exposure gap in early-career work directly. Using the Quarterly Workforce Indicators, which are built from the unemployment-insurance wage records covering a near-universe of private-sector jobs, Tucker found that employment of 22-to-24-year-olds in the most AI-exposed fifth of industry-state cells fell 12% on a regression-adjusted basis over the ten quarters after ChatGPT's release — 15.2% unadjusted between the fourth quarter of 2022 and the second of 2025 — while employment in less exposed industries held steady. He then scaled it: a loss of about 159,000 jobs, or 2.4% of all early-career employment. The declines run deepest in Information (-30%), Professional, Scientific and Technical Services (-15%) and Finance and Insurance (-11%). The mechanism matters more than the magnitude. Hires of early-career workers in the most exposed industries dropped about 9% immediately at ChatGPT's release and have not recovered, running 12.7% below the reference quarters; separations fell too, by 11.5%. A counterfactual decomposition attributes 100% of the employment loss to the hiring shortfall — had hiring held, early-2025 employment would have been unchanged. These are jobs that were never created, not people who were let go, which is why the figure cannot be added to any tally of layoffs. Tucker tests the obvious alternative and largely rules it out: high-exposure industries are not unusually sensitive to monetary-policy shocks, which can account for at most about a quarter of the gap. He flags the same confounders the Stanford work does — the relative shift begins at the COVID onset, and both remote work and educational attainment correlate with AI exposure — and the Census Bureau attaches its standard disclaimer that the paper has not been through the review accorded its official publications. Tucker calls his 12% 'highly consistent with the 16% regression-adjusted relative decline' Brynjolfsson and colleagues reported from private payroll data: two entirely separate datasets, one a selected payroll panel and one a near-universe of federal wage records, arriving at concordant magnitudes and the same mechanism.

A Stanford Institute for Economic Policy Research (SIEPR) policy brief, 'What is really happening to jobs? Separating AI hype from reality,' assembled the fast-growing body of research on AI and work and landed on the reassuring side of it. Its central figure runs directly against the displacement narrative: since 2022, the year ChatGPT launched, the unemployment rate for the fifth of workers most exposed to AI rose by 0.77 percentage points — slightly less than the 0.85-point rise for the least-exposed — which the authors read as a broadly softening labor market rather than one being hollowed out by AI. Employment in highly exposed occupations is 'fairly stable'; online job postings for software developers, one of the most exposed roles, have grown faster than for other occupations over the past year; and firms that adopted enterprise AI grew employment about 10% in the two years after. The one plausible pocket of AI effect the brief flags is the same one every credible source keeps returning to — recent graduates, whose unemployment hit 5.6% in early 2026 — though it notes the Brynjolfsson 'canaries' decline is not clearly notable until 2024 and that remote work and pandemic over-hiring are confounders. What gives the brief its weight is its lead author: Erika McEntarfer served as Commissioner of the Bureau of Labor Statistics until August 2025, the official who ran the very statistics the page's corroborated line waits on. Her verdict, hedged 'early evidence is hardly the last word,' is that AI's aggregate labor-market impact is likely small right now — the corroboration layer speaking in its own voice, and still isolating no headcount of AI-caused cuts.

Montefiore Medical Center in the Bronx sent notices dated 28 May eliminating 12 utilization-review nurse positions — the nurses who read patient charts and argue insurers into covering the care a doctor has ordered — after moving that work onto AI-powered software the union identified as Datavant's. The New York State Nurses Association filed a class-action grievance on 1 June, contending the move breaches a contract clause, won in a strike earlier this year, that requires management to meet the union before AI 'diminishes' union jobs; NYSNA called the cuts AI replacement outright. Montefiore's spokesperson called that characterization 'inaccurate and misleading,' saying only that the system is 'always investing in new technology,' and did not confirm the notices or name the software. Challenger, which tallies AI-cited layoffs, declined to log these as AI at all, recording them instead under 'Technological Update (possibly AI)' — the category it uses when a company cites new technology but does not tie the cuts directly to AI. Small in headcount, it is the first health-care case to fold the page's central ambiguity into a single frame: the same twelve jobs read as AI displacement, as ordinary modernization, or as unprovable, depending entirely on who is asked — and it is the reason the claimed line's own foundation, employers' voluntary attribution, cannot be taken at face value.

Connecticut enacted the AI Responsibility and Transparency Act — the CART Act, Senate Bill 5, Public Act 26-15 — and from October 1, 2026 it requires any employer filing a WARN mass-layoff notice in the state to state whether the layoffs are related to the employer's use of artificial intelligence or another technological change. It is the first US statute to attach an AI-attribution disclosure to a mandatory layoff filing — the exact instrument every cumulative tally of AI-attributed cuts has lacked, since those counts rest entirely on what employers volunteer in press releases rather than on anything they are legally obliged to file. An affirmative answer carries a further consequence: it marks the employer as a 'deployer' of an automated employment decision tool, triggering individualized employee notices from October 1, 2027. The mechanism is a state-level analog to the WARN amendment inside the stalled federal AI Workforce PREPARE Act (S.3339); where the federal bill has had no markup or floor vote, Connecticut's has force of law. The state's 2026 WARN filers to date — Macy's (about 1,000 positions), Aetna (313) and Stanley Black & Decker (about 300) — disclosed no AI cause under the outgoing regime, so the first notices filed under the new duty will be the first real test of whether a statutory question can move the corroborated count off zero. (Sourcing note: the two law-firm analyses on the record differ on the signing date — one dates it to May, the other specifies June 2, 2026, used here as the more specific; the primary bill text at the Connecticut General Assembly could not be retrieved.)

Fortune, drawing on Indeed data, reported UK software-developer postings up 14 percent — concentrated in senior and AI-linked roles — while manufacturing postings have fallen 58 percent since June 2022. The two-speed pattern matches the American evidence on this record: demand rising for the experienced workers who direct AI and falling for both the entry rungs below them and the sectors the technology bypasses. As with the U.S. data, the mechanism is not directly observable from postings — a fallen count conflates AI substitution with everything else that has hit UK manufacturing since 2022 — and the series is a vacancy measure, not an employment one.

Alphabet, Booz Allen Hamilton, CSX and Lattice signaled plans to increase hiring to meet growth targets in AI and cloud businesses, per reporting on U.S. companies stepping up recruitment after an AI-driven pullback. Set against the month's layoff announcements, the hiring column complicates the wipeout narrative this page keeps testing: the same technology cited as the reason for cuts at some employers is the growth engine requiring headcount at others. The record still lacks what it has lacked all year — any economy-wide displacement signal in official statistics — and firm-level announcements in both directions are what fill the vacuum.

German insurer Ergo, a Munich Re subsidiary, announced plans to cut roughly 1,000 positions by 2030 as AI systems take over policy administration and claims-handling tasks. The announcement is notable on this record for its explicitness: where most large employers cutting staff decline to attribute the cuts to AI — the pattern documented here across Visa, Uber and Microsoft in the same month — Ergo named the technology, the functions and the timeline. A four-year horizon on 1,000 roles is attrition-speed automation, not a layoff event, which is itself informative about how a regulated insurer expects the substitution to proceed.

Axios reported OpenAI research finding that nearly half of occupation-specific ChatGPT requests involve work typically handled by a different profession — marketers running analyses, engineers drafting legal language, managers writing code. The cross-occupation pattern cuts against the assumption, built into most exposure studies, that AI's effect on an occupation is bounded by that occupation's own task list: if workers routinely reach across professional boundaries, displacement and augmentation both propagate along paths the occupational statistics do not measure. Company research on the company's own product, with the usual caveat that follows.

A Brookings analysis of career pathways found that AI exposure concentrates in the 'gateway' jobs — administrative support, customer service, junior analysis — that have historically connected lower-wage work to higher-wage careers for the roughly 70 million American workers without a bachelor's degree. The report's claim is structural rather than a layoff count: even where AI eliminates few positions outright, automating the entry rungs erodes the ladder itself, leaving the degree-less workforce with fewer routes upward. It is the clearest institutional statement yet of a pattern this record has tracked piecemeal — entry-level postings falling while senior AI-adjacent hiring rises.

The Senate HELP Subcommittee on Employment and Workplace Safety held a hearing titled 'The Impact of AI on the Workforce,' and its chairman opened by conceding the premise that has made this argument unresolvable. 'Existing labor statistics too often fail to tell us how occupations' tasks are changing, what skills are needed, which jobs will grow and shrink and how workers move through the labor force because of AI,' said Jim Banks of Indiana. Behind the hearing sits S.3339, the AI Workforce PREPARE Act, which Banks introduced on 3 December 2025 with Maggie Hassan, John Hickenlooper and Jon Husted, and which has sat in committee since. It would build an AI Workforce Research Hub at the Labor Department, add AI questions to federal surveys, run a pilot producing statistics on how AI-affected workers change jobs, improve the Bureau of Labor Statistics' occupational projections — and amend the WARN Act so that employees are told whether AI was a substantial factor in a layoff. That last provision is the one that bears on the tally: every cumulative count of AI-attributed job cuts now rests on what employers volunteer in press releases, and a disclosure duty attached to the notices they must already file would put the same claim under statutory obligation. No markup or floor vote has been scheduled. The witnesses were markedly cooler than the legislation's framing. Carol Rogers of the Indiana Business Research Center asked Congress to 'create standards for the definitions so that we compare apples to apples.' Ken Clark of EmployIndy argued that 'the greatest long-term workforce risk is not widespread unemployment. It's the disruption of career pathways that workers rely on to gain experience,' and that AI is reshaping far more jobs than it is replacing. Justin Heck of Opportunity@Work warned that a firm automating the demanding parts of its jobs 'may see short-term efficiency gains while inadvertently dismantling the pipeline that produces its own future supervisors and managers.'

Revelio Labs, a workforce-data firm, launched a monthly US AI Labor Market Tracker written by Lisa K. Simon, Ben Zweig and Caelan Wilkie-Rogers. Its first edition replicates the Stanford-ADP entry-level result on different data — employment in the most AI-exposed occupations has grown about 4% less than in the least-exposed since ChatGPT's launch, with a steeper gap for younger workers — and adds a finding that cuts the other way: firms it identifies as AI adopters have grown headcount 27% more than non-adopters since October 2022, senior roles by 31% against 6% for junior ones. Revelio flags the hole in that number itself, noting that adopting firms were already growing faster before they adopted, so the comparison is not a clean experiment. Its most useful contribution is a mechanism rather than a magnitude. An index of how far the economy's mix of work activities has shifted reached 8.4 percentage points year on year in June 2026 — the share of headcount-weighted activity that would have to be reallocated to return to June 2025's composition — and most of that movement is happening inside occupations rather than between them. If the change is in what a job consists of rather than in how many of that job exist, occupation-level statistics will keep finding very little, which is roughly what they have been finding. Revelio's own summary: 'AI is automating work, but not jobs.' The caveats are the firm's own; it is a commercial data product with unstated provenance for its underlying workforce records and no published methodology for its indices, so it belongs beside the official series rather than in them.

Amazon told staff it was eliminating about 16,000 corporate roles worldwide, its second mass round in three months and its largest since the 27,000 of 2023. As in October, the company did not name AI. Beth Galetti, the senior vice-president who announced it, described the exercise as 'reducing layers, increasing ownership, and removing bureaucracy,' and said Amazon would keep hiring in strategic areas. Andy Jassy had put the same case more bluntly in October: 'It's culture. And if you grow as fast as we did for several years… you end up with a lot more people than what you had before, and you end up with a lot more layers' — a chief executive ruling out the cause his own June 2025 remarks had invited, and both sets of words are his. What the round demonstrates is how thin the documentary record is beneath a headline number. On the day of the announcement California's workforce agency said it had received no WARN notice from Amazon, and the agencies in Washington and Virginia did not immediately report one either; the statutory filings, which are the only legally compelled account of a layoff's size and location, arrived after the press. Washington's eventually documented 2,198 of the cuts in the Seattle area, more than 1,400 in Seattle and about 630 in Bellevue — roughly one in seven of the announced total, with the geography of the rest undisclosed. Amazon's most recent quarter at the time showed profit up nearly 40% to about $21 billion on revenue above $180 billion.

A Financial Times analysis published on 24 July found U.S. technology companies have cut nearly 140,000 jobs since the start of 2026, with Amazon, Oracle, Meta and Microsoft alone accounting for almost 50,000 of them while pouring hundreds of billions into AI data centres. Its most consequential finding for this page is about credibility rather than headcount: companies that cited AI as a factor in their job cuts underperformed the Nasdaq by almost 10% over the 30 trading days after the announcement. Where economists have argued that some employers dress ordinary cost-cutting in AI language, this is a priced, quantitative version of the same skepticism — investors appear to discount the story rather than reward it. The FT also recorded the flow in the other direction: AI-focused companies including Anthropic and OpenAI are hiring rapidly and absorbing talent shed elsewhere, Meta moved roughly 7,000 employees into new AI-focused roles while laying off 8,000, and IBM says it is tripling U.S. entry-level hiring for AI and hybrid-cloud roles. None of the three carries an announced headcount attributed to AI, so none can yet be plotted against the claimed cuts. The FT article is paywalled; these findings are as reported by TechCrunch, which names and links the analysis. (Anthropic built the model that ran this update cycle; the claim about its hiring is the FT's independent reporting, recorded with that attribution and not asserted here.)

The Canaries Dashboard, the Stanford Digital Economy Lab and ADP Research tracker behind this page's one measured AI employment signal, published an update splitting its numbers by gender for the first time. Young women aged 22 to 25 do show weaker employment growth than young men — but the researchers found the gap is not correlated with AI exposure. In the least-exposed fifth of occupations, employment among women in that age band grew 1.3% a year after late 2022 against 2.7% for men; in the most-exposed category, women's employment shrank 4.5% a year against men's 2.5% — a gap of a similar width at both ends of the exposure scale. What drives the headline difference is occupational sorting that predates generative AI: 43.8% of women work in the most-exposed category and 21.2% in the second-most, versus 32.4% and 18.1% of men. "These gaps are a feature of our broader sample; they are not noticeably correlated with AI exposure," the team wrote in materials shared with Fortune. The same update sharpens the finding that does track exposure: the entry-level divergence in the most AI-exposed occupations has not reversed since the original 2025 paper but deepened, by roughly half a percentage point a month. Drawn from 4.6 million workers across more than 730 occupations, with data through June 2026, and, researcher Bharat Chandar told Fortune, robust to controlling for work from home. His caveat stands: "That does not prove that AI is causing these trends, but we think they are important facts to document." Like every corroboration-layer finding on this page, it is a rate, not a headcount — which is why the corroborated line on the graphic above still cannot move.

Challenger, Gray & Christmas reported 45,849 announced U.S. job cuts in June 2026, down 53% from May's 97,006 and 4% below June 2025 — the quietest month since December. Artificial intelligence was again the single most-cited reason, the fourth consecutive month it has led: 14,029 cuts, or 31% of the month's total, ahead of market and economic conditions (12,470) and closings (11,837). AI has now been cited in 101,743 announcements in 2026 alone, about 23% of the year's cuts, and in 173,568 since 2023, when the firm first tracked it as a distinct reason. Technology led all sectors with 15,503 cuts in June and 139,156 for the year, up 83% year over year and nearly a third of all 2026 cuts. Running the other way, and rarely reported alongside the layoff totals: employers announced plans to hire 91,405 workers in the first half of 2026, up 10% on the same period of 2025. "Employers appear to be modestly hiring more workers this year, which would buck the trend since 2020," said Andy Challenger, the firm's chief revenue officer, who also cautioned that the cuts "remain concentrated in technology." The hiring plans carry no AI attribution, so they cannot be set against the AI-cited cuts as a like-for-like offset.

Federal Reserve Bank of St. Louis researchers William Rodgers III and Alice Kassens found the employment rate among 18-to-24-year-olds fell by more than two percentage points between April 2023, the labor market's strongest point, and December 2025 — showing up as higher unemployment rather than people leaving the labor force, "indicating that younger workers were still searching for jobs but with fewer opportunities available." There was no comparable slide for workers aged 25 to 64. Roughly one-third of the increase in 18-to-24 unemployment, they estimated, was attributable to rising demand for the skills AI jobs require; the rest tracks a broad "low-hire, low-fire" decline in openings. AI matters, they wrote, but in a "narrow, early and age-specific way," "raising the bar for young workers trying to secure their first foothold in the labor market." It is the first quantified attribution of a share of a headline labor statistic to AI from inside the Federal Reserve system — and it still yields no headcount of jobs eliminated, because the authors' own scope note says their job-postings measure shows how job content is changing, "not whether firms are using AI to automate jobs or screen applicants." Built on the Current Population Survey and JOLTS. The St. Louis Fed's own post could not be retrieved through this project's fetcher on two separate cycles; the findings here are as reported by CIO Dive, which names the authors, date, method and quotes, and corroborated by Fortune's independent reference to the same paper.

Challenger reported that AI was cited for 38,579 of May 2026's job cuts — 40% of the month's total and the highest monthly AI figure since tracking began, with AI's share climbing from 7% in January to 40% in May and 2026 AI-cited cuts (87,714) already surpassing all of 2025. Yet Challenger's own Andy Challenger cautioned this is 'not yet the jobpocalypse some predicted' and may reflect budgets being 'redirected to AI' rather than jobs literally done by AI, and Apollo's chief economist Torsten Sløk wrote he saw 'zero evidence of job losses because of AI' in payroll data. The claimed line accelerating, the corroborated line still flat — the gap this page tracks, at its widest.

Meta cut about 8,000 employees — roughly 10% of its workforce — with CEO Mark Zuckerberg telling staff 'success isn't a given' as the company reoriented around AI. Meta did not formally attribute the cuts to AI displacement; the AI context is the reorganization it is funding, not a stated cause of the layoffs.

Applying a synthetic difference-in-differences design to CPS microdata, the Budget Lab at Yale found no statistically distinguishable AI effect on either the employment or the inflation-adjusted wages of AI-exposed occupations — 'the estimate is close to zero and cannot be distinguished from it.' It attributed the softening labor market (payroll growth ~20,000 net jobs/month; unemployment up from a 3.4% low in April 2023 to 4.3% in March 2026) to cyclical forces, not AI. The counter-evidence anchor, restated with a stronger method.

Anthropic's labor-market research ranked the 10 most AI-exposed U.S. occupations by the share of tasks AI could speed up or perform — computer programmers 75%, customer-service reps 70%, data-entry keyers 67%, medical-record specialists 67% — while stating plainly that there is 'limited evidence that AI has affected employment to date,' with only 'suggestive evidence' that hiring of younger workers has slowed in exposed roles. A notable admission from a lab whose CEO had forecast mass entry-level displacement ten months earlier: exposure is high, measured impact is not yet.

An IMF Staff Discussion Note estimated that about 60% of jobs in advanced economies are exposed to AI, that roughly 1 in 10 job postings now requires at least one new (largely IT/AI) skill, and that in occupations highly exposed to AI but with limited complementarity, employment ran 3.6% lower five years after those skills appeared in regions with greater AI-skill demand. A measured, if cautious, corroboration that exposure is beginning to translate into relative employment effects for the least-complementary roles.

Federal Reserve Bank of Dallas economists concluded the overall AI impact on employment had been 'small and subtle' as of early 2026, while a companion report found early-career workers' employment did drop in the most AI-exposed occupations — echoing Stanford's cohort finding within an otherwise-quiet aggregate. Another official-research data point that the macro disruption has not arrived even as the entry-level signal recurs.

Challenger, Gray & Christmas reported U.S. employers announced 1,206,374 job cuts in 2025 — up 58% on 2024 and the most since the pandemic year 2020. Employers cited AI as the reason for 54,836 of them (a cumulative 71,825 since Challenger began coding the reason in 2023). But AI accounted for under 4.5% of the year's cuts and did not crack the top five causes; economic conditions (~253,000) and company closings (~191,000) drove far more. The definitive 2025 snapshot of the claimed layer — and of how small AI's share still was even among announcements that named it.

Amazon announced roughly 14,000 corporate-role cuts (about 4% of its corporate workforce), its biggest since 2023, while ramping AI spending. SVP Beth Galetti's message framed the reductions as 'reducing bureaucracy, removing layers, and shifting resources' — pointedly not attributing them to AI, four months after CEO Andy Jassy had said AI would shrink the corporate workforce. A clean example of the gap between AI displacement rhetoric and the reasons companies give on the record.

Yale's nonpartisan Budget Lab found the U.S. occupational mix had not diverged unusually in the 33 months since ChatGPT's release: measured by a dissimilarity index on monthly Current Population Survey data, the AI-era shift was only about 1 percentage point beyond the 1996–2002 internet-adoption period, and exposure/automation/augmentation measures showed no relationship to changes in employment or unemployment — 'the broader labor market has not experienced a discernible disruption.' It did note slight recent upward movement in the early-career gap, consistent with the Stanford finding. The most rigorous counterweight to the displacement narrative.

Two reference points frame why AI-attributed layoffs are so hard to verify. The U.S. Bureau of Labor Statistics incorporates AI into its 2023–33 employment projections only qualitatively — there is no dedicated 'AI' series — and it still projects growth, not decline, for most AI-exposed occupations (software developers +17.9% vs a 4.0% all-occupation average; personal financial advisors +17.1%). Meanwhile the Anthropic Economic Index found AI adoption fast but concentrated: 40% of U.S. employees report using AI at work (up from 20% in 2023), enterprise API usage is 77% automation-oriented versus ~50% for consumer use, and usage per capita is sharply uneven across regions. The gap between adoption and measured job loss is the whole story.

Using monthly ADP payroll records for millions of U.S. workers, Erik Brynjolfsson and colleagues at the Stanford Digital Economy Lab found that workers aged 22–25 in the most AI-exposed occupations — software developers, customer-service reps, accountants — saw a 13% relative employment decline since late 2022 (revised to 16% in the November release, controlling for firm-level shocks), while employment for older workers in the same jobs and for all ages in less-exposed jobs held steady or grew. The first large-scale measured signal that generative AI is reshaping the bottom rung of the job ladder — the strongest single piece of corroboration on this page, though a cohort-specific rate rather than a headcount.

India's Tata Consultancy Services said it would cut about 2% of its workforce (~12,200 roles), mostly middle and senior management, in the biggest layoff in its history. Press coverage framed it as an AI-driven shakeup of the $283-billion IT-outsourcing sector, but TCS attributed the cuts to skills mismatches and deployment/restructuring, not AI — a marquee disputed-attribution case where company and press accounts diverge.

Microsoft laid off about 9,000 employees (under 4% of its workforce), its largest round since 2023, lifting 2025 cuts above 15,000. The company attributed the cuts to organizational streamlining, not AI, even as CEO Satya Nadella said AI now writes 20–30% of Microsoft's code and the firm poured tens of billions into AI infrastructure. A case where the press and analysts supplied the AI narrative that the company itself withheld.