25%
of global employment is in occupations with some exposure to generative AI.
What it does not measure: It does not measure job losses or how far firms adopt the technology; it measures which tasks the technology could theoretically perform.
Numbers · documented reference
A short reference to the key official figures on AI, the labour market, skills, Internet connectivity and cybersecurity. For every figure: what it measures, what it does not measure, its original source with a direct link, its publication date and the date we last checked it.
Last checked: · 21 figures in 5 sections
25%
of global employment is in occupations with some exposure to generative AI.
What it does not measure: It does not measure job losses or how far firms adopt the technology; it measures which tasks the technology could theoretically perform.
34%
potential exposure in high-income countries (against 25% globally).
What it does not measure: An estimate for high-income countries only; it does not apply to every economy.
9.6% vs 3.5%
share of the jobs at highest risk of automation in women’s versus men’s employment in high-income countries.
What it does not measure: Potential exposure, not actual job loss, and specific to high-income countries.
170 million / 92 million
new jobs against displaced jobs expected by 2030, a net increase of 78 million; the new jobs equal about 14% of today’s employment.
What it does not measure: Employer expectations (1,000+ of the largest employers) driven by broad trends: technology, the green transition, and economic and demographic shifts. They are not specific to AI alone.
22%
scale of structural labour-market change (jobs created and destroyed) by 2030, as a share of today’s jobs.
What it does not measure: Not the share of jobs that will disappear: it adds creation and destruction together, and it is a projection, not a measurement.
39%
of key job-market skills that employers expect to change by 2030 (against 44% in the 2023 report).
What it does not measure: Not a share of jobs, but a share of key skills, and an employer expectation.
59 in every 100
workers would need training by 2030, if the world’s workforce were 100 people.
What it does not measure: An estimate built on an employer survey, not a measurement of workers’ actual needs.
63%
of employers name skills gaps as the biggest barrier to transforming their business over 2025–2030.
What it does not measure: The view of surveyed employers, who are among the largest.
50%
of workers have completed training, reskilling or upskilling (against 41% in the 2023 edition).
What it does not measure: It shows how widespread training is, not its quality or its effect on jobs.
Fastest rising
in importance among skills: AI and big data, then networks and cybersecurity, then technological literacy.
What it does not measure: A ranking of skills expected to matter, according to employers, not measured demand in the labour market.
20.0%
of EU enterprises with 10 or more employees used AI technologies in 2025.
What it does not measure: Specific to the EU and to enterprises with 10+ employees in defined sectors; it does not apply to Arab countries or to the world.
13.5% → 20.0%
the rise between 2024 and 2025 (6.5 percentage points), against 8.1% in 2023 and 7.7% in 2021.
What it does not measure: A comparison across years of the same European survey only; do not compare it with figures from other surveys that use different definitions.
6.3 billion (76%)
people use the Internet in 2026, against 6.1 billion (74%) in 2025 on the revised estimate.
What it does not measure: ITU’s estimate of Internet use, not of connection quality or affordability.
About 2 billion
people remain offline in 2026.
What it does not measure: An estimate that is revised each year as new data become available.
93% vs 32%
of the population uses the Internet in high-income countries against low-income countries (2026).
What it does not measure: Figures for income groups under the World Bank classification, not for individual countries.
74%
Internet use in ITU’s Arab States region, against 80% in Asia-Pacific and 42% in Africa (2026).
What it does not measure: ITU’s regional grouping, which may differ from how other sources define “Arab countries”.
41%
Internet use in least developed countries, and 43% in landlocked developing countries (2026).
What it does not measure: Although they are among the fastest-growing (8.7% and 6.5% growth in 2026), they stay well below the global average.
79% vs 74%
of men against women use the Internet worldwide (2026).
What it does not measure: A global average; the gap varies a great deal between countries and regions.
1,008,597
complaints were received by the FBI’s IC3 in 2025, the first time the total passed one million (859,532 in 2024).
What it does not measure: Voluntary complaints from the United States, not a comprehensive measure of cybercrime worldwide.
$20.877 billion
in reported losses in 2025, up 26% on 2024.
What it does not measure: Losses reported by victims in complaints that are not all verified, so actual figures may differ.
22,364
AI-related complaints, with adjusted losses of about $893 million, in the first dedicated AI section of an IC3 report.
What it does not measure: Covers only cases where AI use was identified, and is specific to the United States.
| Date | Change |
|---|---|
| 17 November 2026 | Page published with 21 figures. Latest editions of the sources at publication: ITU Facts and Figures 2026 (3 November 2026), FBI IC3 for 2025 (April 2026), Eurostat for 2025 (December 2025), WEF Future of Jobs 2025 (January 2025) and ILO–NASK (May 2025). |
Eurostat’s 2026 data and a new edition of the WEF Future of Jobs report have not been released yet. We will add them with their dates when they appear.