ETD

• AI Governance
• Artificial Intelligence
• China
• Comparative Analysis
• Democratic Accountability
• European Union (EU)
• State Direction
• Technological Sovereignty

Artificial intelligence challenges public authorities not only because technological development can outpace legislative processes, but also because the capacity to shape AI depends upon resources such as data, computing power, cloud infrastructure, semiconductors and specialised expertise that are distributed across jurisdictions and often controlled by private actors. This dissertation examines how public authorities seek to shape these technological trajectories through a comparative analysis of the European Union and China.

To capture the full range of state intervention, the dissertation develops an analytical framework that distinguishes four overlapping modes of directional governance – strategic, regulatory, infrastructural and social – and argues that directional capacity depends on how these modes are combined rather than on the presence of any single instrument. These modes are evaluated against two normative criteria: technological sovereignty and democratic accountability. Through doctrinal and qualitative comparative analysis of legislation, policy instruments, institutional arrangements and relevant judicial decisions, the dissertation examines how different combinations of legal, strategic, infrastructural and social interventions produce different forms of directional capacity.

The comparison reveals a paradox: the EU and China employ functionally similar governance instruments such as data governance, risk-sensitive regulation, transparency, human oversight and measures addressing the social consequences of AI but organise them through fundamentally different institutional architectures. The EU's regulatory-first trajectory has progressively incorporated strategic and infrastructural direction, while China's strategic-first model has increasingly relied on detailed regulation. Yet neither model is monolithic: the EU's fragmentation is constitutional, dispersing authority and strengthening scrutiny, whereas China's fragmentation is administrative, allowing strategic coordination alongside overlapping mandates and more limited independent contestation.

The normative assessment reveals that technological sovereignty and democratic accountability can coexist. The EU demonstrates stronger procedural accountability and regulatory capacity but remains materially dependent; China demonstrates stronger strategic and infrastructural coordination but weaker independent scrutiny. Both remain constrained by US technological power, showing that domestic governance cannot be assessed in isolation from global technological interdependence.

The conclusion is that directional capacity should be assessed not by the amount of regulation a jurisdiction adopts, but by its ability to connect public objectives with the institutions, resources and mechanisms necessary to shape technological trajectories. Institutional learning is possible, but primarily through adaptation of governance functions rather than imitation of institutional forms.