Key findings

  • Vietnam scores 10.0/10 on risk velocity with an average AI exposure of 3.27/10. Its score rests on 5,734 secure servers per million people; our data has no broadband figure for Vietnam.
  • Nigeria scores 0.1/10 with 0.1 broadband subscriptions per 100 people and 98 secure servers per million. Its workforce of 71.4M has an average exposure of 3.31/10.
  • Saudi Arabia is the clearest mismatch: average exposure of 4.82/10 but risk velocity of only 1.0/10, because it has 483 secure servers per million and no broadband figure in our data.
  • Risk velocity is not a forecast. It is an infrastructure index; it says nothing about when AI will change any job.

What risk velocity actually measures

Most coverage of AI and jobs focuses on exposure: how many tasks in a given occupation can AI replicate? A clerical worker in Nigeria and one in the UK both sit at 8.5/10 on our model-estimated exposure scale, because the score belongs to the occupation group and is the same in every country. What differs between countries is how many workers are in each group and how much digital infrastructure surrounds them.

Risk velocity captures only the second part, and it uses two inputs, both from the World Bank:

They are combined as a geometric mean and scaled from 0 to 10. Where a country has only one of the two indicators, the score uses that one alone. 63 of 198 scored countries have no broadband figure. Earlier versions of this article listed five inputs, including enterprise AI adoption, regulation and education; the index never used them, and that description has been removed.

Why exposure alone is not the whole picture

Two countries can have the same average exposure and very different infrastructure. Risk velocity captures that difference. It does not turn exposure into a timeline: nothing in the index measures employer adoption, cost, regulation or skills.

The top of the ranking: 49 countries at 8.0 or higher

21 countries score the 10.0 maximum, and 49 score 8.0 or higher. The maximum group includes the US (196,554 secure servers per million), the UK, Sweden, Vietnam and South Africa. South Korea scores 9.8, Germany 9.6, the Netherlands 9.3, Japan 8.8 and Australia 8.5.

A caution about the 10.0 group: for 20 of its 21 members our data has no broadband figure, so the score rests on secure-server counts alone. Countries that do have a broadband figure just under the cap, such as Germany (45.6 per 100), score slightly lower for that reason. Part of the gap between 9.6 and 10.0 is therefore a data-coverage effect. See the full list of 21 countries at 10.0.

Vietnam and South Africa show that the top group is not defined by wealth or by high exposure. Vietnam has 5,734 secure servers per million and average exposure of 3.27/10. South Africa has 16,658 per million and GDP per capita of $6,598. The index counts hosting and connectivity, not the nature of the economy.

The middle range: scores between 2 and 8

Brazil scores 6.9 with 24.1 broadband subscriptions per 100 people and 6,941 secure servers per million. China scores 5.2: broadband is high (47.2 per 100) but secure servers are 1,413 per million, well under the cap. Malaysia also scores 5.2, with the opposite pattern (13.5 broadband per 100 and 7,475 servers per million). Mexico (2.2) scores low mainly on secure servers (532 per million) and Indonesia (2.4) mainly on broadband (4.9 per 100 people).

These scores are snapshots of 2022 to 2024 data. They move as connectivity and hosting change, and they say nothing about how quickly any employer in these countries will adopt AI.

The lowest scores

22 countries score 0.0. Among the large economies, Nigeria (0.1), Pakistan (0.3), the Philippines (0.3), India (1.2) and Bangladesh (1.3) score lowest, and together they employ over 700 million workers. Nigeria's 0.1 reflects 0.1 fixed broadband subscriptions per 100 people and 98 secure servers per million.

India (1.2, 476.6M workers) shows the limit of a national index. India has a large technology sector, but ILO data put informal employment at 87.2%, and the score reflects national averages of 3.2 broadband per 100 and 1,212 servers per million, not the situation in individual cities or firms.

Saudi Arabia (1.0) is the most striking mismatch with exposure. Its average exposure of 4.82/10 is similar to developed economies, but the score rests on 483 secure servers per million with no broadband figure in our data. Earlier versions of this article attributed that to expatriate labour shares and regulation; the index has no such inputs, so that explanation was not supported and has been removed.

Risk velocity: country comparison

CountryRisk velocityAvg AI exposureBroadband per 100 peopleSecure servers per 1M peopleWorkers
United Kingdom10.0/105.08/10n/a68,44934.1M
United States10.0/105.06/10n/a196,554144.0M
Vietnam10.0/103.27/10n/a5,73453.7M
South Africa10.0/104.84/10n/a16,65812.7M
Sweden10.0/105.32/10n/a55,0975.3M
South Korea9.8/104.85/1047.811,90228.8M
Germany9.6/105.30/1045.6152,11442.1M
Netherlands9.3/105.44/1043.0198,6829.8M
Australia8.5/104.95/1036.544,29517.0M
Japan8.8/104.92/1038.632,92970.5M
Brazil6.9/104.13/1024.16,941102.1M
China5.2/104.48/1047.21,413362.2M
Malaysia5.2/104.13/1013.57,47513.9M
Mexico2.2/103.83/1021.753259.0M
Indonesia2.4/103.44/104.92,939139.2M
Saudi Arabia1.0/104.82/10n/a48314.2M
India1.2/103.26/103.21,212476.6M
Bangladesh1.3/103.21/108.152769.1M
Pakistan0.3/103.31/101.511077.6M
Philippines0.3/103.62/10n/a13246.9M
Nigeria0.1/103.31/100.19871.4M

Sources: World Bank Open Data (broadband per 100 people and secure servers per million people, 2022-2024), ILO ILOSTAT (employment), WorldJobsData (AI exposure, model-estimated). n/a means no figure in our data. Scores are not forecasts.

Exposure and readiness: why they differ

Vietnam and South Africa both score 10.0. South Africa has 4.84/10 average exposure and 12.7M workers; Vietnam has 3.27/10 and 53.7M. Saudi Arabia has 4.82/10 and scores 1.0. Ranking countries by exposure alone and ranking them by readiness alone give different orders, and neither is a forecast.

The sensible reading is to keep the two questions apart: exposure asks which tasks AI can do, readiness asks whether the basic infrastructure is there. Neither measures how many workers actually use AI. For that, see AI exposure vs AI use by country.

What this means for individual workers

The most useful number for your own situation is the exposure of your occupation, not your country's readiness score. Check it with the job risk calculator. Exposure measures how much of a job's tasks AI can already do; it does not say your job will disappear or when.

Country readiness is context. In a high-readiness country, AI tools are easier to deploy because the basic infrastructure exists. In a low-readiness country, that infrastructure is thin today and may change. Either way, the score cannot tell you what your employer will do.

Check your country's readiness and exposure

Explore AI exposure scores and the risk velocity index for over 200 countries, broken down by occupation group.

Open the explorer →

Was this analysis useful?

Methodology and sources. Risk velocity is built by WorldJobsData from two World Bank indicators: fixed broadband subscriptions per 100 people (IT.NET.BBND.P2, capped at 50) and secure internet servers per million people (IT.NET.SECR.P6, capped at 5,000), combined as a geometric mean and scaled from 0 to 10. Where only one indicator exists for a country, the score uses that one alone. Data years are 2022 to 2024. The index does not use occupations, measured AI adoption or any forecast; the "1-3 years" to "12+ years" wording on some pages is a label for score ranges, not a prediction. AI exposure scores are model-estimated by an AI language model and are not measurements.

AI exposure, robotics, offshoring and work-from-home scores are model-estimated: an AI language model (Claude Opus 4.7) scored each ISCO-08 occupation group on 28 May 2026, guided by published research. They are informed judgements, not measurements, and not forecasts of job losses. "Risk velocity" and its time windows come from a digital-infrastructure index (broadband and secure servers), not from occupations or measured AI adoption. Read the full methodology and how the scores were checked.

Frequently asked questions

What is AI risk velocity?
Risk velocity is a 0 to 10 index of digital infrastructure readiness: fixed broadband per 100 people and secure internet servers per million people (World Bank). It is not a forecast of when AI will affect jobs.
Why does Vietnam score 10.0 on risk velocity?
Vietnam has 5,734 secure servers per million people, above the 5,000 cap, and our data has no broadband figure for it, so the score rests on servers alone. Its average AI exposure is only 3.27/10.
Which countries have the lowest risk velocity?
22 countries score 0.0. Among large economies, Nigeria (0.1), Pakistan (0.3) and the Philippines (0.3) score lowest. Low scores mean little broadband and few secure servers, not low exposure.
Is risk velocity a forecast of AI job losses?
No. The "1-3 years" to "12+ years" wording is a label for score ranges of an infrastructure index. It does not use occupations or measured AI adoption and cannot say when jobs will change.

Spotted an error or know a better source? Report it and we will check it against the source data.