By Samuel Goodger a Policy Analyst at the European Policy Centre.
This blog was produced based on the European Policy Centre policy brief “Fostering AI Resilience in the EU Labour Market”, co-authored in March 2026 by Samuel Goodger, Pietro Valetto and Elizabeth Kuiper.
In Greece and Italy, roughly a third of the workforce may be at risk of displacement from generative AI. Yet, neither country appears among the EU member states most impacted overall — a consequence of lower specialisation in AI-exposed sectors. Among EU firms, generative AI adoption today varies significantly, ranging from 5% in Romania to over 40% in Denmark, mirroring differences in countries’ socio-economic profile. While forecasts of mass disruption have failed to materialise — beyond software engineering — as AI capabilities along the “jagged frontier” continue to evolve, its potential to automate or complement labour warrant close attention.
In a recent EPC analysis, applying the AI Occupational Exposure Index methodology alongside a complementarity measure to EU-SILC microdata,[1] we map workers’ potential to see their jobs displaced (exposure) or augmented (complementarity) by generative AI. Below, we focus on the findings by country and along the income distribution — further investigation of its impacts by gender, age and education level can be found in the full policy brief.
Our analysis reveals a clear split. Wealthy, services-oriented economies have around two-thirds of their workforces in high-exposure occupations, while Eastern European economies have fewer than half of workers exposed.
[1] Insights limited to 26 EU Member States; 2024 data for Hungary was not available at the time of our analysis
Belgium, Denmark and Sweden combine high exposure with a workforce better placed to benefit — more workers hold the complementarity protection that may turn AI into an advantage. Luxembourg, the Netherlands and Ireland present a less favourable picture: high exposure, but more workers at risk of displacement. Most concerningly, only 19% of Greek and Italian workers are set to benefit, with both low complementarity and exposure.
Looking at income, workers in the lower half of the distribution display low exposure and complementarity, with both metrics rising sharply at the top. Worryingly, should middle-income earners — the primary tax base — lose out, this may foster unprecedented inequality amongst white-collar jobs. A narrower tax base would raise redistribution demands just as welfare systems already face strain from demographic change — a recipe for political resentment.
The EU’s current legislative toolkit is ill-equipped to build resilience against these challenges. While the AI Act classifies AI in employment as a high-risk use case, generative AI models fall under a different, lighter regime for General-Purpose AI (GPAI). The Platform Work Directive marked a turning point, but only covers the algorithmic management of gig workers. The Quality Jobs Act is the primary legislative opportunity to close this gap, yet in its Roadmap, the Commission limited new action to “targeted complementary measures”, which may prove inadequate given AI’s potential impacts.
What the evidence demands, and what current frameworks do not yet provide, is a set of targeted interventions spanning workplace governance, skills and fiscal resilience:
- Ensure a strong Quality Jobs Act. The European Commission should introduce expanded employer risk assessments for AI deployment affecting job design, workload and skill requirements, and extend the Platform Work Directive’s transparency and consultation requirements to all sectors.
- Establish quarterly monitoring capacity. The Commission should work with Eurostat and the OECD to track capital-labour substitution signals, wage polarisation, sectoral employment shifts and training uptake, disaggregated by occupation, gender and age.
- Prioritise mid-career resilience and structured entry-level pathways. Re-skilling initiatives programmes should prioritise mid-career workers — in addition to closing persistent gaps in baseline digital skills. Apprenticeship schemes should preserve entry-level pathways, as AI raises expectations. Skills policy should mainstream hybrid intelligence as its guiding framework, combining technical AI literacy with domain expertise and distinctly human capabilities.
- Stress-test welfare systems. The European Commission’s fiscal monitoring bodies should model AI-driven erosion of the tax and unemployment insurance base, feeding findings into European Semester country recommendations and opening structured dialogues on fiscal revenue diversification, including digital services levies and capital income taxes. Well-enforced digital and competition policies to discourage excessive concentration and regulatory capture must accompany these efforts — particularly in light of the EU’s upcoming revamp of merger guidelines.
- Strengthen social dialogue. Drawing on models such as works councils, or tripartite governance, the EU should encourage workplace AI agreements through future Council Recommendations under the European Pillar of Social Rights. Such agreements should cover impact assessments and help identify newly emerging roles. Algorithmic co-design is central to steering technological outcomes toward augmentation rather than substitution.
In addition, the EU’s next budget, the Multiannual Financial Framework, should also be aligned with AI exposure patterns — directing the European Social Fund+ and Just Transition Fund instruments toward regions with the widest preparedness gaps.
These recommendations rest on an evidence base that will need continuous updates. Exposure indices rely on assumptions about AI capabilities that may soon be outdated, and employment effects ultimately depend on real-world adoption, which lags behind technical progress. Diffusion of agentic systems may extend potential displacement risks further into higher-order tasks currently considered safe.
AI is already reshaping which skills are most valued in the workplace, and who benefits from work, with effects concentrated among the groups that have historically financed European welfare systems. The distributional effects of AI adoption are not fixed — they will reflect policy choices for workplace governance, skills investment and welfare resilience, not simply those made by AI’s developers. The EU’s window to shape this transition is open. It will not remain so indefinitely.