The diplomacy of artificial intelligence is becoming, in 2026, one of the most consequential lines of strategic competition among the major powers. Artificial intelligence is no longer merely a technology. It is emerging as an infrastructure of power, comparable to energy, finance, the defence industry, submarine cables, or communications networks. Those who control the models, the chips, the data centres, the cloud, the data flows, and the standards of deployment do not merely control a new market. They control part of the capacity of states, corporations, and institutions to interpret reality, to take decisions, and to act faster than their rivals.
This is the stake that Europe is still struggling to turn from diagnosis into strategy. The European Union today possesses the most advanced regulatory framework for artificial intelligence, yet it has not built a comparable critical mass in infrastructure, capital, foundation models, or scaling capacity. The strategic risk is not that Europe regulates too much. The risk is that Europe becomes the space that writes the rules for a technology developed, financed, and operated elsewhere.
Artificial intelligence is becoming the cognitive infrastructure of power
In the twentieth century, the great powers were defined by their control of maritime routes, energy, heavy industry, currency, and military capability. In the twenty-first century, these criteria do not disappear, but a new layer is added to them: cognitive infrastructure. Artificial intelligence will be used for diplomatic analysis, economic forecasting, military research, cybersecurity, public administration, crisis management, education, healthcare, logistics, and information warfare.
AI is therefore not merely an industry. It is a structural lever of institutional power. States that produce and control it will hold a structural advantage over states that merely import it. The difference will not be solely economic, but diplomatic: some capitals will build the systems through which others take decisions.
In this logic, Europe’s normative advantage is real but insufficient. The European Commission presents the AI Continent Action Plan as a project designed to turn the European Union into a global leader in artificial intelligence, through AI factories, future AI gigafactories, and investment in the order of hundreds of billions of euros. This marks a significant shift in tone: Brussels is no longer speaking solely about rules, but about infrastructure. The question is whether Europe’s pace can keep up with the American and Chinese tempo.
Europe has rules, but not yet enough technological power
The AI Act is, without question, the most significant global attempt to introduce clear rules for artificial intelligence. The official implementation timeline, however, reveals the gradual nature of the European approach. According to the EU AI Act Service Desk, the legislation applies progressively, with full implementation scheduled by 2 August 2027. Prohibitions and AI literacy obligations began to apply in February 2025, obligations for general-purpose AI models entered into force in August 2025, and the majority of rules and enforcement mechanisms become operational from 2 August 2026.
This framework can become an instrument of influence. Europe has previously exported global rules through the size of its market: data protection, competition, environmental standards, product safety. But AI differs from many earlier domains. Here, power does not derive solely from the ability to regulate market access. It derives from the capacity to build the models, to train them, to run them, to integrate them industrially, and to export them.
The data reveal the imbalance. According to the Stanford HAI AI Index 2025, private investment in AI in the United States reached 109.1 billion dollars in 2024, nearly twelve times that of China and twenty-four times that of the United Kingdom. The same report shows that American institutions produced forty notable AI models in 2024, compared with fifteen in China and only three in Europe.
This is the strategic problem: Europe has a sophisticated market, strong universities, technical talent, solid industrial companies, and a powerful legal tradition. Yet in the AI race, these assets must be rapidly converted into infrastructure, capital, and products. Otherwise, regulation remains defensive: it protects the market but does not secure power.
The United States is not selling technology alone. It is selling an ecosystem
Washington treats artificial intelligence as a matter of national security and global leadership. The White House has presented America’s AI Action Plan as a strategy organised around three pillars: accelerating innovation, building American AI infrastructure, and international leadership in diplomacy and security. The document includes more than ninety federal actions and explicitly emphasises the consolidation of American dominance in AI.
This framing is essential. The United States does not seek only to develop competitive AI products. It seeks to impose an ecosystem: chips, cloud, models, applications, standards, research, global companies, and technological alliances. For America’s partners, this ecosystem offers capability, speed, and security. But it also generates dependence.
From a diplomatic perspective, AI is becoming an extension of American power. In the past, Washington built influence through military alliances, financial institutions, trade, technology, and culture. To this architecture, AI infrastructure is now being added. States that adopt American models, cloud services, and platforms at scale connect themselves not only to a technology, but to an order of dependencies, standards, and strategic priorities.
For Europe, this is a delicate dilemma. The United States remains the principal strategic ally, and transatlantic cooperation is vital. Yet a Europe structurally dependent on American AI infrastructure will struggle to speak credibly about digital sovereignty.
China is building a diplomatic alternative
China does not enter the AI competition merely as a technological rival to the United States. It enters it also as a provider of alternative governance. The Ministry of Foreign Affairs of the People’s Republic of China has published the Global AI Governance Action Plan, launched in the context of the World Artificial Intelligence Conference 2025 in Shanghai, with emphasis on international cooperation, global and regional platforms for dialogue, and exchanges among researchers, developers, and regulators.
This line is not new. The Chinese diplomatic mission had already published in 2023 the Global AI Governance Initiative, a document in which Beijing speaks of global collaboration, the healthy development of AI, public access to technologies, and open source.
For the West, these formulations must be read with caution. Analyses published by the Carnegie Endowment for International Peace show that China’s internal AI governance architecture combines selective external openness with a strict system of political control, algorithmic supervision, and alignment of models with the priorities of the Chinese Communist Party. In other words, Beijing’s global diplomatic message on cooperation and open access coexists with a domestic regime that treats AI as an instrument of governance, security, and informational control. This asymmetry between external discourse and internal practice is an essential component of China’s artificial intelligence diplomacy.
From a diplomatic standpoint, however, the message is effective. China presents itself as an alternative to American dominance and as a partner for states that fear the AI revolution will deepen the gap between the leading technological powers and the rest of the world.
This narrative may find traction in the Global South. For many states, the question is not whether the European model is more ethical or whether the American model is more efficient. The question is who offers access, infrastructure, financing, technology transfer, and autonomy of use. It is here that China is seeking to transform AI into an instrument of economic and technological diplomacy.
AI multilateralism: between shared principles and hard competition
Artificial intelligence has already entered the multilateral agenda. The OECD notes that its AI Principles, adopted in 2019 and updated in 2024, seek to promote artificial intelligence that is innovative, trustworthy, and compatible with human rights, safety, and informational integrity. The OECD AI Principles, adopted in 2019 and updated in 2024, are used as a policy benchmark by OECD member states and international partners. Their importance lies less in their legal force than in the fact that they have become the reference used by UNESCO, the G7, the G20, and the Council of Europe to build compatible normative frameworks.
At the global level, Reuters reported that the United Nations General Assembly adopted in March 2024 the first global resolution on artificial intelligence, with an emphasis on protecting personal data, monitoring risks, and respecting human rights. The resolution, initiated by the United States and co-sponsored by more than 120 states, including China, was adopted by consensus, a rare occurrence in a field marked by acute strategic competition. Yet this very consensus underscores the limit: the text sets out broad principles but imposes no binding obligations and does not redress the asymmetry of technological power.
The tension between principles and practice is most visible in the divergence of standards. The United States favours vertical sectoral regulation and industry self-regulation. The European Union imposes a horizontal framework with graduated obligations. China is building a hierarchical system, in which the state sets the limits and the permissible content. All these models invoke the same multilateral concepts — safety, trust, rights, transparency — but operationalise them differently, according to their own strategic interests.
This is the central tension of AI diplomacy: all actors speak of safety, transparency, and cooperation, yet each major power seeks to maximise its own advantage. The United States seeks to preserve technological leadership. China seeks to reduce dependence on the American ecosystem and to offer a global alternative. Europe seeks to defend a model grounded in rights, rules, and democratic oversight. Europe’s problem is that, without sufficient infrastructure of its own, its model risks being legally respected but strategically marginal.
For Romania and Eastern Europe, the risk is cognitive periphery
This debate is not abstract for Romania. If AI becomes critical infrastructure, states that fail to build local or regional capacity will remain consumers of technology. They will integrate models developed elsewhere, depend on foreign cloud services, import solutions for administration, security, education, and the economy, but will contribute little to the definition of standards and technological direction.
For Eastern Europe, the risk is the emergence of a cognitive periphery: regions that use AI intensively but do not produce it, do not control it, and have no leverage over its underlying infrastructure. Cognitive periphery is not measured in GDP or in technological adoption, but in the capacity to influence the architecture of models, the rules governing data training, and the strategic priorities of the platforms used by national administrations, armed forces, and economies. In such a configuration, the divide is no longer only between West and East in terms of capital or industry, but between the centres that generate artificial intelligence and the markets that merely consume it.
Romania need not compete with the United States or China in developing the largest foundation models. But it must define a realistic role for itself: data centres, cybersecurity, industrial applications, AI for public administration, defence, technical education, European interoperability, and strategic analytical capacity. In a regional environment shaped by the war in Ukraine, disinformation, cyber pressure, and military competition, AI is also becoming a matter of national security.
For Romania, the question is not whether AI will matter. The question is whether the state, the universities, and the private sector will act swiftly enough to ensure that Romania does not remain at the margins of Europe’s new infrastructure of power.
Europe must turn regulation into power
Europe need not abandon its normative model. On the contrary, that model may prove a genuine advantage in a world where AI raises real questions of transparency, fundamental rights, safety, accountability, and democratic oversight. But rules cannot replace infrastructure. A Europe that regulates AI without producing it at scale will hold limited influence over the future.
The true stake is not the choice between regulation and innovation. That is a false dilemma. The real stake is the choice between influence and dependence. The United States has understood AI as an ecosystem of technological leadership. China treats it as an instrument of development, internal control, and external influence. Europe seeks to turn it into a domain of trust and rights. For that model to carry weight, it must be supported by computing centres, capital, data, models, talent, and industrial adoption.
Europe risks regulating a future it no longer controls, not because it has too many rules, but because it does not yet have sufficient material power behind them. In the new diplomacy of artificial intelligence, rules will count only if they are matched by capability. Those who control the infrastructure will control the tempo. Those who control the tempo will control the standards. And those who control the standards will hold a decisive diplomatic advantage in the international order now taking shape.
Read also
