Benachrichtigungen
Melden Sie sich hier für E-Mail-Benachrichtigungen zu GIGA-Aktivitäten an
Soziale Medien
Folgen Sie uns
Melden Sie sich hier für E-Mail-Benachrichtigungen zu GIGA-Aktivitäten an
Folgen Sie uns
The conversation around AI often swings between utopian dreams and dystopian fears. As millions of jobs face disruption, the core challenge is not automation itself but the unequal distribution of accompanying benefits. Without broad access to capital and income stability, automation risks deepening inequality rather than expanding opportunity globally.
Middle-income economies like Brazil, China, India, or Indonesia can integrate AI into their existing industrial bases, while many African states seeing informal employment dominate lack AI preparedness. Meanwhile, lower-income countries risk losing service-sector advantages without gaining new AI industries, as automation erodes labour-cost advantages and accelerates “jobless growth.”
AI shifts income from labour to capital. Countries with strong tech ecosystems can accumulate domestic AI capital, while others remain dependent on foreign platforms. This deepens inequality within the Global South, creating a divide between AI adopters and AI-dependent economies.
Evidence points to rising informality and declining labour-market shares. While technology can create jobs, many – especially platform-related roles – remain informal and lack protection, as stable formal employment and effective reskilling systems continue to be scarce.
Less than one-fifth of workers in many regions of the Global South have access to unemployment protection. Without adaptive welfare systems, automation risks pushing millions into precarity and permanent labour-market exclusion.
Effective social protection and large-scale retraining are vital. The EU can promote inclusive digital partnerships, expand AI capacity-building, scale adaptive social safeguards, and reduce technological dependency. That while leveraging Germany’s credibility in data protection, vocational training, and industrial standards to advance trusted, interoperable AI governance globally.
By early 2026, AI had moved from abstract promise to tangible economic, social, and geopolitical reality. Measurable adoption across numerous industries coincided with growing evidence of productivity gains, labour-market disruption, governance challenges, and widening cross-country inequalities, as documented by major international institutions. Importantly, these effects are not evenly distributed but reflect core asymmetries in capital, infrastructure, and skills.
The World Bank (2025) highlighted that over 40 per cent of ChatGPT’s global traffic in mid‑2025 came from middle‑income countries, with Brazil, India, and Indonesia among the top users. China and India already stand out in AI research output and the ability to implement it into industrial systems at scale. This pattern suggests that AI diffusion is most rapid where an existing industrial and human‑capital base allows firms to absorb new technologies more easily.
Countries with robust infrastructure, pre-existing industries, and workforce capacity are more able to harness AI for growth, as illustrated in Figure 1 below. The International Monetary Fund’s AI Preparedness Index (AIPI) assesses AI adoption readiness across 174 countries. It aggregates normalised indicators on digital infrastructure, human capital and labour policies, innovation capacity, and legal frameworks, drawing on data from eight international institutions to provide an indicative, non-ranking measure of preparedness. Under current trajectories, a pronounced AI divide is emerging whereby middle‑ and lower‑income countries – particularly in Africa – lag behind as regards AI adoption, raising the risk that technological change reinforces rather than narrows global inequality.
AI adoption is closely tied to access to electricity, because digital technologies cannot function without dependable sources of power. At the firm level, reliable and affordable electricity lowers the fixed costs of adopting these automated digital systems. In advanced economies, widespread electrification enables firms to integrate AI into manufacturing, services, and everyday life, accelerating productivity gains. By contrast, nearly 775 million people in low‑income countries still lack access to electricity (see Figure 2), which severely limits the ability of workers and businesses to adopt AI tools. This gap becomes especially clear at the technological frontier. A single hyperscale AI project – such as the planned “Stargate” campuses – requires continuous baseload electricity comparable to that of a large power plant, together with advanced grid stability, cooling systems, water access, and high-capacity data infrastructure. The UAE–US AI Campus in Abu Dhabi exemplifies this scale. Developed with the involvement of OpenAI, NVIDIA, Oracle, Cisco, and SoftBank, the project is designed to deliver up to 5 gigawatts of AI data centre capacity to support hyperscale training and inference workloads. Initial phases, reportedly beginning with several hundred megawatts and scaling toward a first 1 GW cluster later in the decade, highlight how frontier AI development increasingly depends on energy and infrastructure conditions that remain out of reach for most low-income economies.
This uneven access to electricity underscores a broader pattern: AI adoption under prevailing ownership and governance models tends to widen income inequalities across countries, increasing the gap between advanced and developing economies. It drives a reallocation of jobs away from traditional labour-intensive sectors towards research and development as well as innovation-heavy activities, where advanced economies hold a structural advantage. As a result, developing nations face a heightened risk of falling further behind, as AI reduces opportunities for technological “catch-up” through imitation and late industrialisation. Rather than eliminating jobs outright, AI primarily reshapes their distribution, amplifying disparities across countries and sectors. In plain language, a customer-service centre may hire fewer entry-level agents as chatbots handle basic queries instead, while demand rises for a smaller number of higher-skilled roles such as AI supervisors, data analysts, and system designers. The result is not fewer jobs overall, but a reorientation towards more skilled and, more importantly, less accessible positions.
Capital’s share of global income has shifted significantly over the past four decades, rising from approximately 39 per cent in 1980 to nearly 47 per cent by 2025; labour's equivalent figures, meanwhile, are 61 per cent and 53 per cent respectively. These structural changes reflect globalisation, financial liberalisation, technological advancement, and weakened bargaining power on the part of labour, all of which favour capital owners over wage earners. These dynamics are not inevitable market outcomes but the result of policy and institutional choices that have served to reinforce inequality worldwide over the course of time (Chancel et al. 2026).
These distributional trends are increasingly being reproduced within the AI economy itself. One prominent example is “data annotation,” a critical input for AI systems. The term refers to the process of labelling text, image, audio, or video inputs to aid AI systems’ learning of pattern recognition. Although this work is central to the latter’s development, it is outsourced to the Global South at very low wages, offering few opportunities for skill accumulation or upwards mobility. Those in question often engage in data annotation out of necessity rather than choice, yet the meaning of this work varies per local socio-economic conditions. The same platform-based tasks can function as a primary livelihood for highly educated workers facing macro-economic collapse, as a supplementary income source for women navigating informal labour markets, or as low-mobility employment mediated by small subcontracting firms embedded in transnational value chains. In high-income contexts, similar tasks are more often taken up as residual, flexible work by women balancing paid employment with caregiving responsibilities, highlighting how AI labour reproduces existing inequalities rather than offering uniform opportunities for social advancement.
Similar patterns of precarity extend beyond AI-specific tasks to platform work more broadly. Taken together, these arrangements form a global value chain in which risk and insecurity are decentralised while control and profits remain concentrated. Delivery workers for platforms such as Grab or Uber are usually classified as informal labour because they are engaged via contracts that fail to provide social protection, health insurance, pensions, or union representation. Even if they receive digital payments and sometimes tax IDs, the absence of stable contracts and benefits keeps their jobs informal in nature. Only recently, the International Labour Conference has taken a major step towards rectification by approving the drafting of a binding global “Convention,” alongside a “Recommendation,” to set international standards for platform workers. This initiative aims to secure decent working conditions and fundamental rights for the over 1.5 million people worldwide engaged in platform work, addressing issues such as misclassification, lack of social protection, and transparency in AI-powered algorithmic management.
Moreover, data availability itself reinforces global asymmetries. By using the CC-MAIN-2026-08 snapshot from Common Crawl (2026), Figure 3 shows how AI systems are disproportionately trained on English-language and Global North-generated data, giving advanced economies structural advantages in the development and application of these automated tools. Common Crawl is an open, regularly updated dataset that systematically crawls and archives a large portion of the public web, capturing billions of web pages, including text, metadata, and links. It is widely used in AI research to train and benchmark language models because it reflects the real-world distribution of online content at scale.
This linguistic dominance shapes not only market power but also determines whose knowledge, needs, and social realities are even encoded into AI systems. As a result, low‑income countries find themselves further marginalised. This simultaneously serves to limit the local relevance of AI while strengthening the competitive position of already advanced economies.
Global labour markets are under strain, with nearly 58 per cent of all workers having been employed informally as of 2024. Simultaneously, labour’s share of global income has, as noted, steadily declined, reinforcing conditions that resemble post‑work pressures, where jobs exist but are increasingly insecure, low‑quality, and disconnected from productivity gains. Informal employment is impacting over two billion workers, with nearly 90 per cent of employment being informal in the world’s least developed countries and sub-Saharan Africa (United Nations 2025). At the time of writing, Africa is home to around 1.5 to 1.6 billion people, making it the world's second-most populous continent; it is also the one experiencing the fastest growth in population size globally. However, human capital-rich Africa still struggles economically, as a significant share of its workforce remains in informal employment. This is especially true in Central and West Africa, where agriculture and small-scale trade dominate. Moreover, as seen in Figure 4, Africa has the largest proportion of citizens employed in low-skilled occupations (57 per cent) and the lowest percentage of high-skilled jobs (10 per cent) anywhere in the world. These occupational structures heighten vulnerability to automation without generating compensatory pathways as regards upwards mobility.
AI is increasingly reducing opportunities for low-skilled workers. Research shows a strong link between rising AI investment and declining employment in routine and manual jobs: as AI adoption grows, human involvement in these tasks tends to disappear. Statistical evidence further suggests that AI investment actively drives this shift, underscoring technology’s role in reshaping employment rather than merely responding to it (Giwa and Ngepah 2024).
This displacement, however, does not affect all economies in the same way. IMF findings suggest that about 40 per cent of jobs worldwide are vulnerable to AI. Advanced economies face the highest exposure here (around 60 per cent of jobs), but they are also better positioned to capture productivity gains and create new opportunities (Cazzaniga et al. 2024). Emerging and low-income countries face fewer immediate disruptions, yet they risk falling behind as AI adoption accelerates elsewhere and adjustment capacity remains limited. The International Labour Organization, for instance, warns that service-sector jobs are particularly susceptible to AI takeover, especially roles involving routine or repetitive tasks such as customer service, clerical work, and administrative duties. In countries like the Philippines, where over 61 per cent of workers are employed in services and business-process outsourcing contributes around 7.4 per cent of gross domestic product, a significant share of call-centre jobs are classified as “high exposure” and “low complementarity,” meaning generative AI is more likely to replace than augment human-performed tasks (Cucio and Hennig 2025).
Firm-level evidence from millions of US payroll records shows that younger workers in AI‑exposed jobs are encountering a waning of opportunities (hiring chances), while older employees remain largely unaffected (Brynjolfsson, Chandar, and Chen 2025). A complementary study using résumé and job-posting data from nearly 285,000 firms confirms that those adopting generative AI sharply reduce junior hiring but maintain or expand senior roles, demonstrating how even in advanced economies AI adoption erodes entry‑level pathways and reshapes career ladders (Hosseini Maasoum and Lichtinger 2025).
As digital automation spreads, productivity rises faster than employment creation, a dynamic already observed in advanced economies and highlighted by Goldman Sachs as the risk of so-called “jobless growth” (Hatzius et al. 2025). Additionally, the ILO revised its 2025 forecast, lowering the global job growth rate from 1.7 to 1.5 per cent accordingly and predicting seven million fewer jobs worldwide than previously projected. This occurs because automation erodes labour‑cost advantages and slows hiring, meaning output expands while the number of jobs created and maintained fails to keep pace. This divergence reflects a structural decoupling of output growth from labour demand.
In the Global South, low‑income countries dependent on routine service sectors are particularly vulnerable as AI undermines their cheap‑labour advantage. Where informal employment dominates, automation bypasses workers entirely. These economies are left, in consequence, with rising output but stagnant or declining employment opportunities.
Informal labour is prone to disruption, and those engaged in it lack visibility in official safety nets. The ILO reports that only 18 per cent of unemployed workers worldwide receive related benefits, with coverage particularly low in Africa, Asia, and Latin America. In sub‑Saharan Africa, fewer than 8 per cent of unemployed workers are covered (Cannaday, Barrett, and Zeidan 2025). AI pioneers such as Sam Altman and Elon Musk support a Universal Basic Income to mitigate AI-driven job displacement, but coverage and implementation remain limited. For example, consider an Uber driver who might one day be replaced by autonomous delivery drones from companies like Amazon. This worker typically has no formal contract, little savings, and no access to unemployment benefits or training programmes. There is often no time, money, or structured support to learn new skills or switch to a different type of work, leaving them highly exposed if technology displaces their job.
Part of the problem is that adult learning and reskilling systems often fail to reach those who need them most. According to OECD Skills Outlook (2025), participation in training remains uneven and weak, particularly for low-skilled, older, and displaced workers – the very groups most vulnerable to AI. These programmes are frequently fragmented, with low uptake where it is most needed. Even where training does exist, it is often misaligned with labour-market demands, failing to equip workers with the digital and AI-relevant skills required in rapidly evolving jobs.
Some countries are beginning to explore coordinated strategies to improve access to AI and digital skills, but implementation remains modest, particularly for informal and low‑skilled workers in regions like sub‑Saharan Africa, Latin America, and Southeast Asia. Several African states have adopted national digital transformation strategies that include AI skills development and life-long learning components, and programmes like Nigeria’s “3 Million Technical Talent (3MTT)” initiative aim to train millions in AI and related digital competencies. In Latin America, the likes of Chile and Colombia have launched national AI or digital strategy plans that emphasise talent development and workforce readiness. By contrast, countries with weaker institutions and constrained digital ecosystems – such as Venezuela, where technological infrastructure lags behind and formal AI governance frameworks are only nascent – lack coordinated upskilling or national AI strategies, leaving informal and low‑skilled workers highly exposed to said job shifts.
Such digital capacities are already proving to be a huge help to humanity, saving lives and expanding access to essential services – especially in contexts where infrastructure and resources are limited. In India, for instance, AI-powered tools assist low-literacy rural users in navigating government procedures, accessing healthcare information, and even receiving early warnings about outbreaks of disease. In Kenya and South Africa, AI-driven mobile advisory systems guide smallholder farmers on planting, irrigation, and pest control, improving food security and livelihoods for thousands. In Latin America, AI-enhanced chatbots and educational platforms provide children and young adults in underserved regions with learning opportunities despite teacher shortages. During the COVID-19 pandemic, AI models in countries across Southeast Asia helped predict outbreak hotspots, optimise hospital resource allocation, and accelerate vaccine distribution, with direct benefits for citizens’ well-being. These examples illustrate how AI’s potential extends far beyond commercial profit alone: it can empower vulnerable populations, reduce inequalities in access to information and services, and address urgent humanitarian needs, demonstrating its transformative impact on the everyday experiences of millions of people.
Nonetheless, across all geographic contexts, effective social safety nets and large-scale retraining programmes are essential if vulnerable workers are to be protected and the gains from AI shared more broadly and equitably. From a diplomatic and development-policy perspective, advanced economies – particularly the EU’s – have a strategic role to play in shaping a more inclusive global AI transition:
Digital diplomacy should move beyond market access and regulatory export towards genuine digital partnerships with the Global South. This includes co-investment in infrastructure, open and interoperable AI systems, and shared standards that enable local innovation rather than lock-in to proprietary technologies.
The EU should expand AI capacity-building initiatives as a core pillar of its development cooperation. This entails supporting education and training in data science, AI engineering, and digital governance; strengthening local research institutions; and enabling technology transfers so that firms and public institutions in developing countries can adapt AI to local socio-economic needs. Capacity-building should prioritise AI’s public sector usage (administration, agriculture, energy, healthcare), where related productivity gains can translate into broad-based welfare improvements rather than narrow private rents.
Adaptive social-protection systems must be scaled up to address AI-driven disruption, particularly in economies with high levels of informality to their labour markets. Development cooperation can support the expansion of portable social benefits, unemployment assistance for non-standard workers, and digital delivery systems that increase coverage and responsiveness. Rather than treating social protection as a residual safety net, it should be understood as productive infrastructure enabling workers to adapt, retrain, and to participate in coming digital changes.
Reducing technological dependency is essential for long-term development and geopolitical stability. This requires diversifying data sources, supporting local data ecosystems, strengthening regulatory capacity over data ownership and cross-border data flows, as well as promoting multilingual and locally relevant AI models. Without such measures, AI risks entrenching new forms of reliance in which value creation, decision-making power, and future growth remain concentrated in the hands of a small number of advanced economies and firms.
Development-orientated AI diplomacy can, then, help Global South countries secure a fairer share of related productivity gains while protecting the world’s majority from rising exclusion, precarity, and political marginalisation. The issue here is not whether digital automation will transform global economies, but ensuring that process is governed in ways conducive to shared prosperity rather than heightened inequality.
