AI is changing work everywhere, but each world region tells a different story about it. This page shows how the EU, the German-speaking countries, the US, East Asia and the Global South see and govern the issue. It also shows what people in different countries fear.
Broadly, five narratives can be distinguished. The EU treats AI at work mainly as a question of rights and procedures. The German-speaking countries rely on further training and social partnership. At federal level the US thinks in terms of competition. China sees AI as an engine of growth and employment under state monitoring. In the Global South the digital divide dominates: it holds back the benefits but offers no protection against displacement.
This grouping is our own simplification, not an official typology. At global level, though, the narratives meet at one point: international organisations currently expect jobs to be reshaped rather than to disappear. Whether that holds remains open.
The International Labour Organization estimates that about one in four workers worldwide is in an occupation with some degree of exposure to generative AI. According to the ILO, most jobs are more likely to be transformed than made redundant. Media and web occupations are now more exposed too, because AI can generate images, voices and video.
An advisory body to the UN Secretary-General presented the report 'Governing AI for Humanity' with seven recommendations in September 2024. As a follow-up, the General Assembly set up an independent international scientific panel on AI and a global dialogue on AI governance. The UN page itself does not treat work as a separate topic.
Pope Leo XIV devoted his first encyclical, 'Magnifica Humanitas', to AI and human dignity. He signed it on 15 May 2026, the 135th anniversary of the labour encyclical 'Rerum Novarum'. It ties work closely to identity. AI should support workers rather than deskill or surveil them.
Source: Georgetown University, 2026
The EU asks less how many jobs will disappear and more how AI decisions about people can stay fair and reviewable. Under the AI Act, AI used for hiring, promotion, dismissal or performance monitoring counts as high-risk. After a delay, however, the related obligations only apply from 2 December 2027.
In parallel there is a debate about dedicated rules for algorithmic management. Parliament and the unions are pushing for them, while employers are sceptical. The Commission is planning a 'Quality Jobs Act' for this. At the time of writing (October 2026) we had not seen an adopted proposal.
The so-called AI Omnibus (Regulation (EU) 2026/1744, in force since 27 July 2026) moved the obligations for stand-alone high-risk uses from 2 August 2026 to 2 December 2027. These include AI in employment. It nevertheless remains classified as high-risk. Most transparency obligations, such as disclosing that people are interacting with an AI, still apply from 2 August 2026.
On 17 December 2025 the European Parliament adopted a resolution on digitalisation, AI and algorithmic management in the workplace, by 451 votes to 45 with 153 abstentions. It calls for limits on purely algorithmic hiring decisions and a right for workers to know what data are collected. Data on emotional states, private conversations and off-duty time should not be processed.
Source: HowTheyVote.eu, 2025
According to the European Parliament's research service, the Quality Jobs Act is scheduled in the Commission's work programme for the fourth quarter of 2026. It is meant to support AI uptake, protect workers from the risks of algorithmic management and clarify how the AI Act and the General Data Protection Regulation interact. The European Trade Union Confederation wants a 'human in control' principle, while the employers' federation BusinessEurope warns against over-regulation. The Platform Work Directive (EU) 2024/2831 has already been adopted.
Source: EPRS briefing, 2026
In an open letter of 13 October 2025 the European Trade Union Confederation (ETUC) called for a dedicated directive on algorithmic management and AI at work. Its demands include human oversight of all AI-driven decisions, consultation and bargaining rights on data and surveillance, and paid AI training. Displacement and deskilling should also be prevented.
Source: ETUC, 2025
In the German-speaking countries AI is mainly seen as a training challenge. The labour market is expected to adapt through further training and co-determination. Income decoupled from paid work plays hardly any role in official policy. Switzerland even held a vote on it.
For a website rooted in Linz this is an interesting contrast: the question of a society with much less paid work has so far barely been raised in politics here.
In June 2023 Johannes Kopf, head of Austria's public employment service AMS, expected no mass unemployment from AI, but very significant effects. Those most affected would be medium- and higher-skilled workers doing routine tasks, for example in offices, text production and legal work. Some jobs would disappear and new ones appear. He sees an enormous need for further training.
Source: ORF, 2023
In the 2025 government programme, the Vienna Chamber of Labour (AK Wien) welcomes the goal of trustworthy and ethical AI and more resources for AI oversight and employee data protection. Among its criticisms: according to the AK, the AMS budget for 2026 is still 100 million euros below that of 2024. It also misses a right to training during working hours.
Source: AK Wien, assessment of the 2025 government programme
When presenting its AI goals in November 2018, the then federal government stated that a basic income would not be part of its AI strategy. It intended to respond to labour-market effects with education measures.
Source: Trending Topics, 2018
According to the Institute for Employment Research (IAB), machines or software could already perform an average of 55 per cent of occupations' core tasks in 2022, 14.5 percentage points more than in 2013. AI is seen as the main driver of the latest increase. The figures rest on expert ratings of tasks. It is unclear whether generative AI is already fully captured. Technically possible also does not mean that jobs actually disappear.
The popular initiative 'For an unconditional basic income' was put to the vote on 5 June 2016 and rejected. Before that, the National Council had turned it down by 157 votes to 19 with 16 abstentions, and the Council of States by 40 to 1 with 3 abstentions.
Source: admin.ch, 2016
At federal level the US focuses on the race for technological leadership. Protection for workers has so far come mostly from individual states. It is precisely these state laws that the Trump administration now wants to push back.
Unions are therefore demanding above all a say in the workplace. This is the counter-model to the EU's horizontal, rights-based approach.
An executive order signed on 11 December 2025 sets up a task force in the Department of Justice to challenge state AI laws. States with 'onerous' rules are to lose certain funding. Colorado's AI law, which also covers discrimination in employment decisions, is named explicitly. The law firm Gibson Dunn expects only limited legal effect in the short term.
Source: Gibson Dunn, 2025
The union federation AFL-CIO launched an initiative on AI on 15 October 2025. It aims to limit AI surveillance and unjust AI-driven layoffs, and calls for human oversight of employment decisions and for workers to be involved in development and deployment. Transparency and retraining are also on the list.
Source: SHRM, 2025
China explicitly builds AI into its employment strategy and wants the state to monitor the effects. East Asia is also the centre of industrial robotics: most robots are installed here, and robot density is highest here.
South Korea has had its own AI framework act since January 2026. We have not yet been able to check Japan's official AI strategy against original sources. What stands out there is how little people worry about their own jobs (see below).
In June 2026 the State Council presented an employment plan running to 2030. Unlike the previous plan, which treated AI mainly as a risk, it makes AI part of economic and employment strategy. It provides for new occupations, a programme for retraining and transitions, and a system to monitor job creation and displacement by AI. The target is surveyed urban unemployment below 5.5 per cent.
Source: Caixin Global, 2026
In 2024, 542,000 industrial robots were installed worldwide. About 54 per cent of them, 295,000 units, went to China, a record. Japan followed with 44,500, the US with 34,200 and South Korea with 30,600 units. Germany stood at just under 27,000.
Source: IFR World Robotics, 2025
In 2024 there was a global average of 132 robots per 10,000 manufacturing workers. South Korea leads with 1,220, Germany stands at 449, Japan at 446, the US at 307 and the EU at 231. China reached 166, an increase of 17 per cent. The release gives no figure for Austria.
Source: IFR, robot density, 2026
South Korea's framework act on the development of AI and the establishment of trust has applied since 22 January 2026. It is aimed mainly at businesses that provide AI. Employers may be covered if AI materially influences workplace decisions. Enforcement is phased and initially focuses on guidance.
Source: vLex, employer analysis, 2026
In the Global South, AI mainly affects formal, digitally connected jobs: IT and outsourced office services in India, clerical work in Latin America, call centres in the Gulf. These very jobs long served as a ladder for advancement. At the same time, many workers who could benefit from AI lack reliable internet access.
We have not yet checked any original source on Africa. We are therefore deliberately leaving the continent open rather than asserting facts we cannot back up. That does not mean there is no debate there.
A roadmap by the government think tank NITI Aayog with the IT industry body NASSCOM and the consultancy BCG (October 2025) estimates that without countermeasures the tech sector could lose more than 1.5 million jobs by 2031. With timely action, up to 4 million jobs could be created. It proposes an 'India AI Talent Mission', AI skills in education and large-scale retraining. These are scenarios, not measurements.
Source: Outlook Money, 2025
According to an ILO and World Bank study (2024), 26 to 38 per cent of jobs in Latin America and the Caribbean are exposed to generative AI. 8 to 14 per cent could become more productive, while 2 to 5 per cent face full automation. Up to half of the productivity potential is blocked by the digital divide. Low digitalisation offers no protection against automation.
A paper by the ILO and the World Bank from March 2026 examines 135 countries. In rich countries more jobs are exposed, mainly clerical and professional ones. In poorer countries the workers at risk are often already online, so losses can come quickly. Those who could benefit from AI, by contrast, frequently lack reliable internet.
Source: ILO/World Bank, 2026
Since 2021 Saudi Arabia has required remote customer-service jobs to be held by citizens. Precisely such jobs are highly exposed to AI. The Saudi AI authority SDAIA aims to train one million people in AI, and around 300,000 had enrolled according to the report (as of August 2025).
Source: AGBI, 2025
The African Union adopted its Continental AI Strategy in Accra in July 2024. It sees AI as a chance for new jobs and businesses, especially for women and young people, but also lists job displacement through automation as a risk.
The strategy contains no figures on the effects on Africa's labour markets. It states that these effects are not yet well understood in the African context and calls for more data and research, as well as investment in young people's digital and AI skills.
A report by the African Development Bank (December 2025) estimates that broad, inclusive use of AI could add up to 1 trillion US dollars to GDP by 2035, roughly a third of the continent's current economic output. One reason given is Africa's young population.
The bank's report page gives no concrete figures on jobs created or lost. It is a scenario of potential, not a measurement.
Source: African Development Bank, Africa's AI Productivity Gain, 2025
How far AI can make work more productive depends on basic infrastructure. According to the World Bank (World Development Report 2026, August 2026), nearly a third of rural schools in Sub-Saharan Africa lack reliable electricity and more than two-thirds lack dependable internet access. With partners, the World Bank aims to provide energy access to 300 million people in the region by 2030.
Worry about one's own job is unevenly spread. It is greatest in emerging economies of South-East Asia and smallest in the US and Japan. At the same time, people in the US are the most nervous about AI products. Europeans mostly see AI at work positively, but want it carefully managed.
Surveys measure moods, not actual job losses. But they show the expectations with which societies enter the transition.
In the Ipsos AI Monitor 2025 (30 countries), an average of 35 per cent thought it likely that AI would replace their current job within five years. The figure was highest in Thailand (74 per cent), Malaysia (63), Indonesia (51) and Singapore (50), low in Japan (29) and lowest in the US (26). On average 53 per cent were nervous about AI products, 64 per cent in the US and 29 per cent in Japan. Some country values were read off a chart.
Source: Ipsos AI Monitor, 2025
In a Pew survey of US workers (October 2024, published February 2025), 52 per cent were worried about AI's future impact on the workplace and 36 per cent hopeful. 32 per cent expected fewer job opportunities for themselves, only 6 per cent more. Just 16 per cent did some of their work with AI.
Source: Pew Research Center, 2025
According to a Eurobarometer survey (published February 2025), more than 60 per cent of Europeans view robots and AI at work positively, and more than 70 per cent expect higher productivity. 84 per cent say AI at work needs careful management to protect privacy and transparency.