The European Commission’s Joint Research Centre (JRC) has published a significant new study:

Generative Artificial Intelligence in Secondary Education – Uses and Perceptions from the Perspective of Early Adopters across Five EU Member States (9 December 2025).

While the report focuses on secondary education, its findings are far broader. Education, after all, is one of the most complex and risk-sensitive domains of the Public Sector (PS): it combines regulation, public procurement, data protection, ethics, workforce transformation, and citizen trust. As such, it acts as a proxy for AI readiness across Public Administration.

The study analyses early GenAI adoption across Finland, Germany, Ireland, Luxembourg, and Spain, offering a valuable lens on how different governance models shape AI uptake, constraints, and trajectories.

JRC144345_01

Why These Five Countries Matter

The selection was deliberate. The JRC aimed to cover:

• different administrative traditions

• varying degrees of centralisation

• diverse digital maturity levels

• contrasting policy–implementation gaps

Together, these five countries form a mini-map of Europe’s public-sector diversity.

Mapping the Five Member States to Public-Sector Governance Models

🇫🇮 Finland – Centralised, Trust-Based, Capability-Driven

Finland represents the Nordic governance model:

• high institutional trust

• strong central guidance

• long-term investment in digital skills

AI adoption here is cautious but principled. The emphasis is not on speed, but on AI literacy, ethics, and human agency. This mirrors Finland’s broader public-sector approach: enable innovation through competence rather than outsourcing responsibility.

Pattern: Slow–steady adoption, high sustainability.

🇩🇪 Germany – Federal, Fragmented, Profession-Led

Germany exemplifies a federal and legally conservative model:

• strong Länder autonomy

• uneven implementation

• limited national coordination

Interestingly, teachers and institutions often move faster than policy, creating what the report describes as a “lawless space”: high experimentation, low formal protection. This is typical of German public administration, where innovation often precedes regulation.

Pattern: Bottom-up innovation, delayed harmonisation.

🇮🇪 Ireland – Agile, Central Strategy with Operational Gaps

Ireland reflects an agile, strategy-driven Anglo-Saxon model:

• strong national AI strategy

• openness to experimentation

• reliance on guidance rather than mandates

In education, as in wider PA, AI adoption is led by motivated early adopters, while institutional capacity building is still catching up.

Pattern: Fast strategic positioning, uneven execution.

🇱🇺 Luxembourg – Small State, High Coordination

Luxembourg shows the advantages of scale and coordination:

• centralised decision-making

• strong alignment between strategy and execution

• early curriculum integration

This mirrors Luxembourg’s broader digital government approach: fewer actors, faster alignment, and pragmatic experimentation.

Pattern: High coherence, limited scalability lessons.

🇪🇸 Spain – Decentralised, Policy-Heavy, Implementation-Light

Spain illustrates a highly decentralised model:

• strong national frameworks

• wide regional autonomy

• uneven local capacity

Digital competence is formally embedded in policy, but GenAI adoption varies widely across regions. This is consistent with Spain’s broader public-sector AI landscape, where innovation often depends on regional leadership rather than national push.

Pattern: Strong norms, fragmented delivery.

What About France and Italy?

🇫🇷 France – Centralised Strategy, State-Led AI

France is absent from the sample, but highly relevant.

France follows a state-centric, technocratic model:

• strong central AI strategy

• major role of public research institutions

• preference for sovereign platforms and national champions

In education and public services, France tends to move top-down: pilots, then scale. The risk is slower grassroots adoption; the benefit is strong legal and ethical alignment.

Likely pattern: Structured adoption, strong compliance, slower experimentation.

🇮🇹 Italy – Fragmented Capacity, Strong Normative Pressure

Italy’s case is particularly instructive.

Italy combines:

• strong national regulation

• weak local execution capacity

• high dependence on vendors and system integrators

In education and PA more broadly, AI adoption is often driven by EU funding cycles (PNRR) rather than internal transformation strategies. Skills shortages and organisational inertia remain major constraints.

Likely pattern: Compliance-driven adoption, uneven impact.

What This Tells Us About AI in the European Public Sector

Across all cases, one message is clear:

AI adoption in Public Administration is less about technology and more about governance capacity.

The same GenAI tools produce radically different outcomes depending on:

• administrative culture

• decision-making autonomy

• skills investment

• trust in institutions

• clarity of responsibility

Education, as shown by the JRC study, acts as an early warning system for the broader Public Sector: where education struggles with AI governance, the rest of PA will follow.

Why This Paper Matters Beyond Education

This JRC report is not just about schools. It is about:

• AI literacy as a public-sector capability

• Human-in-the-loop governance

• The gap between policy ambition and operational reality

• The risk of informal, unprotected AI use inside public institutions

For policymakers, CIOs, and public-sector leaders, the message is simple:

AI in government cannot be delegated to tools, vendors, or pilots alone.
It requires institutional design, skills, and accountability.