The U.S. Census Bureau’s September 2026 working paper, “Graduating into Disruption: Labor Market Outcomes for AI-Exposed College Majors,” reports weaker early-career outcomes for graduates from the most AI-exposed college majors. The paper says the most AI-exposed decile of majors saw initial employment fall by 5 percentage points and full-quarter initial earnings fall by 13%.
The paper says those effects began immediately after ChatGPT was introduced in late 2022. In the broader context the supplied sources provide, the Federal Reserve’s July 2026 AI labor note says the labor market shows “more immediate but uncertain impacts” of AI, while the Fed’s 2026 household-employment report says one-in-four workers used generative AI at work in the prior month.
Read together, the supplied sources point to a pattern that is narrower than an across-the-board employment shock. The Census paper focuses on the point of labor-market entry for AI-exposed majors; the Federal Reserve notes describe broader AI use at work and caution that the labor-market effects are still uncertain. The sources do not identify a universal schedule for all graduates, and they do not provide a company-level or sector-wide count of jobs gained or lost.
What the Census paper says
The Census Bureau working paper is titled “Graduating into Disruption: Labor Market Outcomes for AI-Exposed College Majors.” It is identified in the supplied material as a September 2026 working paper and as a primary source released on 2026-09-10. According to the paper, the most AI-exposed decile of college majors saw initial employment fall by 5 percentage points and full-quarter initial earnings fall by 13%.
The supplied research bundle says the paper finds these effects were concentrated at labor-market entry. It also says the divergence began immediately after ChatGPT was introduced in late 2022. The bundle does not identify the majors included in the most AI-exposed decile, and it does not supply a list of individual colleges, employers, or occupations.
The paper’s findings matter because they attach specific labor-market changes to the earliest stage of graduate employment, rather than to later career progression. But the supplied sources do not state whether the paper measures applications, interviews, offer acceptance, first-day employment, or another specific hiring step. The sources also do not say that AI itself caused the reported outcomes; they say the paper links the outcomes to AI exposure and to the period after ChatGPT’s introduction.
How the Federal Reserve context fits
The Federal Reserve Board’s July 2026 note, “The AI Buildout and the Economy: Publicly Available Data to Assess AI’s Impact,” says the labor market shows “more immediate but uncertain impacts” of AI. The supplied bundle says that note provides context for interpreting AI investment and labor-market effects. That framing is narrower than a claim that AI has already transformed the entire labor market, and the supplied sources do not broaden it beyond the wording quoted here.
The Fed’s 2026 household-employment report, “Employment and Job Quality,” says one-in-four workers used generative AI at work in the prior month. The supplied material says that report offers broader context on AI use at work and worker perceptions. It does not specify which industries, job levels, or tasks those workers were using generative AI for, and the supplied sources do not identify whether that use was voluntary, employer-directed, or part of any particular AI deployment program.
That distinction matters because the supplied evidence supports two different kinds of claims: wide workplace use of generative AI on one hand, and measurable early-career labor-market effects among the most exposed majors on the other. The sources do not state that the one-in-four figure and the Census results describe the same population, the same time window, or the same labor-market mechanism.
Who is affected and what changed
Based on the supplied sources, the affected group is graduates from the most AI-exposed college majors, especially at the moment they enter the labor market. The measured change is a 5 percentage point drop in initial employment and a 13% drop in full-quarter initial earnings for the most AI-exposed decile of majors. The supplied bundle does not identify how many graduates are in that decile, so it does not support any estimate of the number of people affected.
The sources also do not state that broader unemployment rose, that all graduates faced lower wages, or that the labor market experienced a general collapse. Instead, the supplied material explicitly says the sharpest hit is at initial labor-market entry and that broader labor-market measures remain mixed. That means the clearest verified signal in the bundle is a hiring and earnings setback for new graduates in exposed fields, not a universal employment downturn.
The supplied sources do not identify whether these effects are temporary, whether they persist after the first job, or whether they vary by geography, employer size, or degree level. They also do not say whether the reported changes reflect direct substitution by AI, changed employer screening, weaker demand in specific fields, or another mechanism. Those causal questions remain outside the scope of the supplied evidence.
What to watch next
Because the Census paper is a working paper, the next question is how its findings hold up as more labor-market data accumulate and as the paper is discussed alongside other evidence. The supplied bundle does not identify a revision schedule, a peer-review timetable, or a follow-up release date. It also does not say whether the Federal Reserve will update either of the cited notes with new estimates.
For now, the bundle supports a cautious reading: AI’s labor-market effects are already visible in early-career outcomes for some college majors, while the broader data still show uncertainty about the size and timing of the shift. The sources do not support a claim that AI has already produced a single, economy-wide employment outcome. They do support the more specific conclusion that the first place to look is hiring and earnings at labor-market entry.
That is the main development to track from here. If future data continue to show that AI-exposed majors are entering weaker first jobs while broader unemployment remains mixed, the policy and business debate will likely stay focused on entry-level pipelines, wage setting, and the transition from college to work. The supplied sources do not go further than that, and this article does not either.
Source note
The factual claims in this article are limited to the supplied research bundle: the Census Bureau working paper, the Federal Reserve Board’s July 2026 AI labor note, and the Federal Reserve Board’s 2026 household-employment report.
Sources
- Graduating into Disruption: Labor Market Outcomes for AI-Exposed College Majors — U.S. Census Bureau
- The AI Buildout and the Economy: Publicly Available Data to Assess AI's Impact — Federal Reserve Board
- Employment and Job Quality — Federal Reserve Board