AI Safety Tracker

AI adoption

33 economies 50 activity classes 2025 survey

Two questions that are usually asked separately. Eurostat asks enterprises what they use. Microsoft estimates what populations use. Setting them side by side shows something neither gives alone: in most of Europe, people are using AI well ahead of the organisations that employ them.

Large EU enterprises using at least one AI technology EU27, enterprises with 250 or more people employed, 2025 survey wave. Use of artificial intelligence in enterprises Eurostat Licence EC-reuse-2011-833-EU Updated annual Retrieved , today Self-reported by enterprises in the ICT usage survey, so it measures what firms say they use and what they recognise as AI. Covers the EU plus a few reporting non-members, and only enterprises with ten or more people employed unless the size dimension says otherwise. Agriculture, forestry, fishing, mining and the financial sector are outside the survey population entirely. Small cells are suppressed rather than zero, and NACE activity classes nest, so shares by industry do not partition anything. Cite Eurostat, "Use of artificial intelligence in enterprises". https://ec.europa.eu/eurostat/web/digital-economy-and-society/database. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/.

55%

Against 17 percent of firms with 10 to 49 people employed. Adoption scales with headcount more cleanly than with anything else here.

All EU enterprises with ten or more staff using AI Use of artificial intelligence in enterprises Eurostat Licence EC-reuse-2011-833-EU Updated annual Retrieved , today Self-reported by enterprises in the ICT usage survey, so it measures what firms say they use and what they recognise as AI. Covers the EU plus a few reporting non-members, and only enterprises with ten or more people employed unless the size dimension says otherwise. Agriculture, forestry, fishing, mining and the financial sector are outside the survey population entirely. Small cells are suppressed rather than zero, and NACE activity classes nest, so shares by industry do not partition anything. Cite Eurostat, "Use of artificial intelligence in enterprises". https://ec.europa.eu/eurostat/web/digital-economy-and-society/database. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/.

19.9%

Ranging from Denmark at 42 percent to Romania at 5.2 percent.

Economies where the population is ahead of its employers Share of working-age population using AI tools, minus share of enterprises with ten or more staff using at least one AI technology. Use of artificial intelligence in enterprises Eurostat Licence EC-reuse-2011-833-EU Updated annual Retrieved , today Self-reported by enterprises in the ICT usage survey, so it measures what firms say they use and what they recognise as AI. Covers the EU plus a few reporting non-members, and only enterprises with ten or more people employed unless the size dimension says otherwise. Agriculture, forestry, fishing, mining and the financial sector are outside the survey population entirely. Small cells are suppressed rather than zero, and NACE activity classes nest, so shares by industry do not partition anything. Cite Eurostat, "Use of artificial intelligence in enterprises". https://ec.europa.eu/eurostat/web/digital-economy-and-society/database. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/. AI Diffusion Report: AI User Share Microsoft AI for Good Lab Licence MIT Updated quarterly Retrieved , today Derived from anonymised Microsoft telemetry, then adjusted for OS/device market share, internet penetration and a modelled mobile-to-desktop ratio. It therefore measures Microsoft-visible AI use, scaled, not observed total AI use. Systematically less reliable where Microsoft platform share is low, and it cannot see AI embedded in products users do not recognise as AI. Cite Microsoft AI for Good Lab, "AI Diffusion Report: AI User Share". https://github.com/microsoft/ai-diffusion-report. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/.

25of 27

Median gap +12 points. Two different surveys asking two different questions, so read the direction, not the size.

Most used AI technology in EU enterprises Use of artificial intelligence in enterprises Eurostat Licence EC-reuse-2011-833-EU Updated annual Retrieved , today Self-reported by enterprises in the ICT usage survey, so it measures what firms say they use and what they recognise as AI. Covers the EU plus a few reporting non-members, and only enterprises with ten or more people employed unless the size dimension says otherwise. Agriculture, forestry, fishing, mining and the financial sector are outside the survey population entirely. Small cells are suppressed rather than zero, and NACE activity classes nest, so shares by industry do not partition anything. Cite Eurostat, "Use of artificial intelligence in enterprises". https://ec.europa.eu/eurostat/web/digital-economy-and-society/database. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/.

Text mining

11.8 percent of firms. Shares overlap: a firm using three is counted in all three.

  • Text or speech generation 8.8%
  • Speech recognition 7.2%
  • Workflow automation or decision support 5.3%
  • Machine learning for data analysis 5.1%

Highest adopting activity: computer programming, consultancy, and information service activities EU27, enterprises with ten or more people employed, 2025. Use of artificial intelligence in enterprises Eurostat Licence EC-reuse-2011-833-EU Updated annual Retrieved , today Self-reported by enterprises in the ICT usage survey, so it measures what firms say they use and what they recognise as AI. Covers the EU plus a few reporting non-members, and only enterprises with ten or more people employed unless the size dimension says otherwise. Agriculture, forestry, fishing, mining and the financial sector are outside the survey population entirely. Small cells are suppressed rather than zero, and NACE activity classes nest, so shares by industry do not partition anything. Cite Eurostat, "Use of artificial intelligence in enterprises". https://ec.europa.eu/eurostat/web/digital-economy-and-society/database. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/.

65.7%

Against 12 percent in manufacture of textiles, wearing apparel, leather and related products.

  • Publishing, motion picture, video, television programme production; sound recording, programming and broadcasting activities 56.2%
  • Scientific research and development 52.1%
  • Advertising and market research; other professional, scientific and technical activities; veterinary activities 48%
  • Manufacture of basic pharmaceutical products and pharmaceutical preparations 41.3%

People against their employers

Share of the Working-age population The denominator used for the population adoption figures, as defined by the publisher of that dataset. Adoption shares on this site are shares of that population, not of total population. Glossary entry using AI tools, against the share of enterprises doing the same. 25 of 27 economies have the population ahead.

Economy People Enterprises Gap
France 49.6% 18.2% +31
Ireland 49.9% 19.6% +30
Spain 45.1% 20.3% +25
Poland 32.3% 8.4% +24
Hungary 33.3% 10.4% +23
Bulgaria 30.7% 8.6% +22
Norway 49.4% 28.9% +21
Portugal 27.4% 11.5% +16
Serbia 25.3% 10.1% +15
Italy 31.4% 16.4% +15
Czechia 31.2% 17.6% +14
Romania 18.0% 5.2% +13
Greece 21.5% 8.9% +13
Bosnia and Herzegovina 23.2% 10.8% +12
Croatia 27.2% 15.2% +12
Türkiye 18.9% 7.4% +11
Netherlands 43.5% 33.2% +10
Albania 19.2% 9.0% +10
Slovakia 27.1% 18.0% +9
Slovenia 30.1% 21.6% +9
Germany 32.3% 26.0% +6
Belgium 40.3% 34.5% +6
Austria 35.4% 29.9% +6
Lithuania 25.1% 21.3% +4
Sweden 37.3% 35.0% +2
Finland 30.6% 37.8% -7
Denmark 32.4% 42.0% -10
Population Enterprises Population figures are 2026-Q2; enterprise figures are the 2025 survey wave. Use of artificial intelligence in enterprises Eurostat Licence EC-reuse-2011-833-EU Updated annual Retrieved , today Self-reported by enterprises in the ICT usage survey, so it measures what firms say they use and what they recognise as AI. Covers the EU plus a few reporting non-members, and only enterprises with ten or more people employed unless the size dimension says otherwise. Agriculture, forestry, fishing, mining and the financial sector are outside the survey population entirely. Small cells are suppressed rather than zero, and NACE activity classes nest, so shares by industry do not partition anything. Cite Eurostat, "Use of artificial intelligence in enterprises". https://ec.europa.eu/eurostat/web/digital-economy-and-society/database. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/. AI Diffusion Report: AI User Share Microsoft AI for Good Lab Licence MIT Updated quarterly Retrieved , today Derived from anonymised Microsoft telemetry, then adjusted for OS/device market share, internet penetration and a modelled mobile-to-desktop ratio. It therefore measures Microsoft-visible AI use, scaled, not observed total AI use. Systematically less reliable where Microsoft platform share is low, and it cannot see AI embedded in products users do not recognise as AI. Cite Microsoft AI for Good Lab, "AI Diffusion Report: AI User Share". https://github.com/microsoft/ai-diffusion-report. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/.

These are two surveys asking two different questions, and the gap between them is not a single measured quantity. Eurostat asks a firm whether it uses any of seven named technologies, which a firm can answer "no" to while its staff use a chatbot daily. Microsoft estimates individual use from its own telemetry, scaled by a model. The direction of the gap is worth something; its exact size is not. Both exclude firms with fewer than ten people employed, where most of Europe actually works.

By number of people employed

Share of enterprises using at least one AI technology, by size band.

10 to 49 staff 17%
50 to 249 staff 30.4%
250 or more staff 55%

Eurostat also publishes overlapping roll-up classes, 0 to 9 and 10 to 249, which are excluded here: plotting a bar next to the bars it contains counts the same firms twice. Firms with fewer than ten people employed are outside the survey population, so the smallest employers in Europe are missing from every figure on this page. Use of artificial intelligence in enterprises Eurostat Licence EC-reuse-2011-833-EU Updated annual Retrieved , today Self-reported by enterprises in the ICT usage survey, so it measures what firms say they use and what they recognise as AI. Covers the EU plus a few reporting non-members, and only enterprises with ten or more people employed unless the size dimension says otherwise. Agriculture, forestry, fishing, mining and the financial sector are outside the survey population entirely. Small cells are suppressed rather than zero, and NACE activity classes nest, so shares by industry do not partition anything. Cite Eurostat, "Use of artificial intelligence in enterprises". https://ec.europa.eu/eurostat/web/digital-economy-and-society/database. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/.

By what the firm does

Share of enterprises using at least one AI technology, by economic activity.

Highest adoption top 14
Computer programming, consultancy, and information service activities 65.7%
Publishing, motion picture, video, television programme production; sound recording, programming and broadcasting activities 56.2%
Scientific research and development 52.1%
Advertising and market research; other professional, scientific and technical activities; veterinary activities 48%
Manufacture of basic pharmaceutical products and pharmaceutical preparations 41.3%
Telecommunications 39.8%
Legal and accounting activities; activities of head offices; management consultancy activities; architectural and engineering activities; technical testing and analysis 37.9%
Manufacture of computer, electronic and optical products 37.4%
Travel agency, tour operator and other reservation service and related activities 35%
Electricity, gas, steam and air conditioning supply 33.6%
Manufacture of coke and refined petroleum products 30.4%
Manufacture of chemicals and chemical products 28.4%
Manufacture of electrical equipment 25.9%
Manufacture of electrical equipment, machinery and equipment n.e.c. 25.6%
Lowest adoption bottom 6
Manufacture of textiles, wearing apparel, leather and related products 12%
Manufacture of food products; beverages and tobacco products 12.5%
Manufacture of products based on: food, beverages, tobacco, textile, leather, wood, pulp and paper; publishing and printing 13.2%
Wholesale and retail trade and repair of motor vehicles and motorcycles 14.3%
Manufacture of basic metals and fabricated metal products, except machinery and equipment 14.3%
Retail trade, except of motor vehicles and motorcycles 15.5%

NACE activity classes nest: broad groupings and their own components both appear in Eurostat's table, so these are a lookup rather than a partition and the shares sum to nothing meaningful. The aggregate rows and the broadest groupings are filtered out here so a category is not ranked against its own contents. Use of artificial intelligence in enterprises Eurostat Licence EC-reuse-2011-833-EU Updated annual Retrieved , today Self-reported by enterprises in the ICT usage survey, so it measures what firms say they use and what they recognise as AI. Covers the EU plus a few reporting non-members, and only enterprises with ten or more people employed unless the size dimension says otherwise. Agriculture, forestry, fishing, mining and the financial sector are outside the survey population entirely. Small cells are suppressed rather than zero, and NACE activity classes nest, so shares by industry do not partition anything. Cite Eurostat, "Use of artificial intelligence in enterprises". https://ec.europa.eu/eurostat/web/digital-economy-and-society/database. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/.

What they use it for

The seven technologies the survey asks about individually, as a share of all enterprises.

Text mining Enterprises using AI technologies performing analysis of written language (text mining) 11.8%
Text or speech generation Enterprises using AI technologies generating written or spoken language or programming codes (natural language generation, speech synthesis) 8.8%
Speech recognition Enterprises using AI technologies converting spoken language into machine-readable format (speech recognition) 7.2%
Workflow automation or decision support Enterprises using AI technologies automating different workflows or assisting in decision making (AI based software robotic process automation) 5.3%
Machine learning for data analysis Enterprises using machine learning (e.g. deep learning) for data analysis 5.1%
Image recognition Enterprises using AI technologies identifying objects or persons based on images (image recognition, image processing) 3.8%
Autonomous robots and vehicles Enterprises using AI technologies enabling physical movement of machines via autonomous decisions based on observation of surroundings (autonomous robots, self-driving vehicles, autonomous drones) 1.4%

Shares overlap and do not sum to the headline rate: a firm using three of these is counted in all three. Labels are shortened for reading; Eurostat's full wording is under each one. The summary indicators Eurostat also publishes, for "at least one", "at least two" and "none", are omitted because they are totals of these seven. Use of artificial intelligence in enterprises Eurostat Licence EC-reuse-2011-833-EU Updated annual Retrieved , today Self-reported by enterprises in the ICT usage survey, so it measures what firms say they use and what they recognise as AI. Covers the EU plus a few reporting non-members, and only enterprises with ten or more people employed unless the size dimension says otherwise. Agriculture, forestry, fishing, mining and the financial sector are outside the survey population entirely. Small cells are suppressed rather than zero, and NACE activity classes nest, so shares by industry do not partition anything. Cite Eurostat, "Use of artificial intelligence in enterprises". https://ec.europa.eu/eurostat/web/digital-economy-and-society/database. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/.

By country

Enterprises with ten or more people employed, using at least one AI technology, 2025.

Country Enterprises using AI
Denmark 42%
Finland 37.8%
Sweden 35%
Belgium 34.5%
Luxembourg 33.6%
Netherlands 33.2%
Austria 29.9%
Norway 28.9%
Germany 26%
Estonia 23.4%
Show the remaining 23
Slovenia 21.6%
Malta 21.5%
Lithuania 21.3%
Spain 20.3%
Ireland 19.6%
France 18.2%
Slovakia 18%
Czechia 17.6%
Italy 16.4%
Croatia 15.2%
Latvia 12.2%
Portugal 11.5%
Bosnia and Herzegovina 10.8%
Hungary 10.4%
Montenegro 10.1%
Serbia 10.1%
Cyprus 9.3%
Albania 9%
Greece 8.9%
Bulgaria 8.6%
Poland 8.4%
Türkiye 7.4%
Romania 5.2%

The EU aggregate and the euro area are excluded from this list, because they are made of the countries in it. A few reporting non-members are included, which is Eurostat's own survey population rather than a choice of ours. Use of artificial intelligence in enterprises Eurostat Licence EC-reuse-2011-833-EU Updated annual Retrieved , today Self-reported by enterprises in the ICT usage survey, so it measures what firms say they use and what they recognise as AI. Covers the EU plus a few reporting non-members, and only enterprises with ten or more people employed unless the size dimension says otherwise. Agriculture, forestry, fishing, mining and the financial sector are outside the survey population entirely. Small cells are suppressed rather than zero, and NACE activity classes nest, so shares by industry do not partition anything. Cite Eurostat, "Use of artificial intelligence in enterprises". https://ec.europa.eu/eurostat/web/digital-economy-and-society/database. Retrieved 28 October 2026 via AI Safety Tracker, https://aisafetytracker.org/adoption/.

Related

  • Usage Token volumes and user counts as companies state them.
  • Divergence map AI use against governance readiness, by economy.
  • Law and policy The instruments that apply to AI in eleven jurisdictions.