AI job-risk watch · checked 31 October 2026
Young workers in exposed U.S. jobs saw another payroll decline
September’s Stanford–ADP update shows employment down 3.7% from a year earlier for ages 22–25 in highly AI-exposed occupations, versus 0.2% growth in the least-exposed. It does not show that AI caused those losses.
Norway: a broader counter-signal
Norwegian employer registers through July 2026 show no clear employment gap between the most and least AI-exposed private-sector occupations across ages 21–60: +1.5 percentage points, standard error 2.7. For ages 21–30 the gap is −0.8 points, standard error 4.0. The score measures task exposure, not actual employer AI use; it cannot cancel the narrower U.S. payroll warning.
Spanish unemployment
A new Spanish occupation study finds 1.2% more registered unemployed workers per 10-point higher AI exposure in its first post-ChatGPT period. The later 1.7% estimate is imprecise; new contracts do not show a persistent decline. The score is not Spanish employer use, and the design cannot prove AI caused the difference. A separate Spanish employment study finds growth across exposure groups through 2025, using another measure and outcome.
U.S. early-career watch
In the selected ADP payroll panel, high-exposure employment fell 0.5% year over year across ages versus 1.0% growth in least-exposed work. For ages 26–30, the split was −2.8% versus +0.9%. The panel is not nationally representative; its occupation score measures possible AI reach, not an employer’s use.
A four-country study finds that remote-work exposure weakens the GenAI link to lower junior hiring shares through 2025. French permanent-hiring records show a later 21–29-year-old entry gap but also cannot isolate AI. German full-time records through 2025 show only a small, statistically uncertain adjusted young-worker exposure link. Yale’s broad U.S. tracker finds no clear economy-wide AI footprint. The sources count different outcomes and populations.
New: firms report little net change, but expect churn
A revised seven-country executive survey reports essentially no net AI-related firm employment change over the past three years. More than 40% expect their own firm’s employment to rise or fall because of AI in the next three. The past estimate is self-reported and the future one is a forecast; neither cancels the young-worker payroll warning.
A gap within exposed occupations
Pew’s U.S. employment analysis finds slower growth for women than men as occupation AI exposure rises, especially among workers without a bachelor’s degree. It compares employment before and after generative AI became widely available; it does not show AI caused the gap.
Job ads describe more AI-related work
An October U.S. working-paper abstract finds that newly classified duties rose from under 1% to nearly 3% of job-ad statements in occupations using AI most between 2022 and 2026. Those ads cannot show that new jobs were created or that workers earned more.
New: a hiring judgment, not a hiring count
A U.S. experiment found that when hiring professionals read an AI-as-substitute warning, they rated fictional new graduates less favorably for interviews and salary talks. That message changed stated judgments; the study did not observe real callbacks or pay.
The answer in one minute
Entry routes may be narrowing in exposed work, but no source here isolates how many jobs AI removed. Remote work, changing demand and employer choices complicate the attribution. Neither a weak AI coefficient in one study nor a strong French hiring gap predicts your own outcome.
Before acting on a job-loss claim
- Did it count hires, employment or layoffs—or only a changing share?
- Did it observe workplace AI use, or infer it from an occupation score?
- Did it compare remote work and other concurrent changes?
A population result cannot show that buying a course will protect your job.