Who Controls AI? The New Battle for Digital Sovereignty

Artificial intelligence is forcing governments to confront a question that traditional ideas of sovereignty were never designed to answer: how can a country claim meaningful control over its digital future when the data, computing power, cloud infrastructure and AI models it depends on may all be controlled elsewhere? The answer increasingly matters not only for cybersecurity, but also for economic development, democratic governance and geopolitical power.

A new study, "Artificial Intelligence and the Reconceptualization of Digital Sovereignty: A Political–Philosophical Analysis," published in the journal Philosophies by Tran Quoc Hung of Ho Chi Minh City University of Banking, argues that AI has made classical territorial sovereignty conceptually inadequate. The paper compares three dominant approaches, the US liberal-market model, the European Union's techno-legal rights model and China's cyber-sovereignty model, and concludes that none fully resolves the tensions created by global technological dependence.

The analysis ultimately proposes a different idea: relational digital sovereignty. Instead of imagining sovereignty as technological self-sufficiency, it argues that meaningful digital autonomy depends on whether countries and citizens can participate fairly in the global systems that govern data, infrastructure, algorithms and knowledge.

AI Has Turned Sovereignty Into an Infrastructure Question

Classical sovereignty rests on a relatively straightforward assumption: a government exercises supreme authority within a defined territory. AI complicates that model because the infrastructure of digital power rarely respects national borders. Data may be generated domestically but stored abroad, processed through foreign cloud platforms and analyzed by models developed by multinational companies.

The paper separates digital sovereignty into several layers, including physical infrastructure such as chips, cables and data centers; software and technical standards; data ownership and circulation; and access to AI models, cloud services and knowledge. A country may therefore possess strong legal control in one layer while remaining deeply dependent in another.

The distinction exposes what the study describes as a form of formal sovereignty without substantive autonomy. A government can require local data storage, regulate automated decision-making and establish national AI rules, yet still rely on foreign processors, foreign cloud platforms or foreign foundation models to make those rules operational. Legal authority alone does not create technological capacity.

This matters most for countries outside the small group that dominates advanced computing. The study highlights the concentration of AI supercomputing, cloud services, semiconductors and models among a handful of states and corporations, arguing that such concentration creates structural disadvantages for smaller economies.

The development-policy implication is profound. Digital sovereignty is no longer simply a cybersecurity or data-protection issue; it is increasingly connected to industrial policy, infrastructure finance, competition policy, education, language technology and access to computing power. Countries that fail to build capacity in these areas may find themselves legally sovereign but technologically subordinate.

The US, EU and China Offer Three Models

The study describes the United States as operating primarily through a liberal-market model centered on private innovation, competition and national security. Its strength lies in technological dynamism, but the analysis identifies a contradiction between support for open information flows and the use of export controls, extraterritorial legal authority and restrictions on foreign technologies when strategic interests are involved.

The European Union follows a markedly different path. Its approach emphasizes fundamental rights, privacy, transparency and legal accountability, with digital sovereignty pursued through regulation rather than direct state control of information. The paper presents this as the most normatively sophisticated of the three models, particularly because it attempts to protect individual autonomy against both governments and corporations.

However, the European model also reveals the limits of law. Strong data-protection rules cannot eliminate dependency on foreign cloud providers or digital infrastructure. The study points to this gap between regulatory ambition and material capacity, arguing that rights become less meaningful if governments lack the infrastructure required to enforce them independently.

China's cyber-sovereignty model sits at the opposite end of the spectrum. It combines data localization, domestic digital platforms and strong state authority over information flows. This can produce greater technical autonomy, the study argues, but at a potentially substantial cost to freedom of information, civic accountability and protection against state surveillance.

Taken together, the comparison highlights a key policy dilemma. Markets can concentrate digital power in private hands; rights-based regulation can become ineffective without infrastructure; and state-centered technological autonomy can undermine individual freedoms. Digital sovereignty therefore cannot be reduced to choosing one of these existing models.

Vietnam Shows Why the Dilemma Is Hardest for Emerging Economies

Vietnam provides the study's most development-relevant case because it sits between competing models. It is a fast-digitalizing, middle-income economy that seeks greater control over domestic data and technology while remaining dependent on foreign investment, international supply chains and external technological expertise.

The paper describes Vietnam's emerging framework as a hybrid. Data-localization requirements and state supervision reflect elements of China's approach, while newer privacy, transparency and AI-governance provisions incorporate aspects of the European rights model. Vietnamese-language AI development is also framed as a form of cultural and linguistic self-determination.

However, the country's economic strategy depends heavily on access to foreign capital, cloud infrastructure, advanced chips and international technology partnerships. Stronger localization rules may increase formal control while simultaneously raising costs, discouraging investment or limiting access to technologies needed to develop a competitive domestic AI sector.

This is the paradox confronting many developing countries. Excessive dependence can leave them vulnerable to decisions made by foreign firms and governments, but excessive technological nationalism may isolate them from the very networks required for upgrading. The study characterizes Vietnam's emerging model as a form of negotiated dependency rather than full technological autonomy.

The same tension is likely to confront economies across Asia, Africa and Latin America. For Global South governments, the challenge is not to replicate the United States, European Union or China, but to secure enough bargaining power, infrastructure and institutional capacity to avoid permanent dependence while preserving access to international markets and innovation.

The Next Sovereignty Debate Will Be About Fair Interdependence

The study proposes to replace the search for technological self-sufficiency with a concept of relational sovereignty. Under this view, autonomy does not require countries to produce every chip, operate every cloud platform or train every AI model domestically. It requires them to participate in digital relationships on terms that do not systematically strip them of agency.

It reframes global AI inequality as more than an investment gap. Access to computing infrastructure, technical standards, data value and AI knowledge becomes a question of international justice. At the same time, sovereignty must protect citizens as well as states; a government cannot claim digital independence while denying its own population informational freedom or meaningful control over personal data.

For policymakers, this points toward a broader agenda. Developing countries may need regional cloud capacity, shared computing infrastructure, stronger negotiating blocs, open technical ecosystems and investment in domestic language models and digital skills. International organizations, meanwhile, face growing pressure to ensure that AI governance does not simply formalize the advantages of countries that already control the infrastructure.

The study is conceptual rather than empirical, and no new dataset is analyzed, so its claims should be read as a framework for interpreting policy choices rather than evidence that one model produces superior outcomes. Yet the framework captures an increasingly visible reality: AI power depends on who controls the systems beneath regulation.

The defining sovereignty question of the AI era may therefore be less about whether countries can disconnect from global technology networks and more about whether they can remain connected without becoming permanently dependent. For much of the developing world, that distinction could determine who shapes the digital future, and who merely consumes it.