AI, Structural Power and Governance Fragmentation: A Systematic Review and the Geopolitical AI Power Framework

When we think of artificial intelligence, the conversation often feels like a high-stakes race between two giants: the United States and China. We hear about who has the fastest computer chips, the most powerful software models, and the brightest engineers. This view explains the competition at the very top, but it leaves out what happens to everyone else. It misses the reality for countries that are not trying to win the race, but are instead being pulled into a system they did not design. These nations find themselves dependent on digital tools, rules, and data flows controlled by others. This is not just about technology; it is about a quiet shift in power. Some countries hold the capacity to set the conditions under which others must operate, while others are left to follow the rules. This dynamic is reshaping the global political economy, creating a situation where adopting new technology can accidentally lock a country into a position of permanent dependence.

A new study by Alfredo Merlet, a researcher at the Universidad de Sevilla, investigates exactly how this happens. The paper asks a simple but profound question: how is artificial intelligence changing the balance of power between nations, and why is the system for governing this technology breaking apart instead of coming together? To find the answer, the author did not just read a few articles; they conducted a massive, systematic review of the existing research. They searched two major scientific databases for thousands of records published between 2018 and 2026. After removing duplicates and filtering out studies that did not fit the criteria, they were left with 330 relevant works. From this large group, they selected a core set of 49 studies to read in depth, analyzing how different experts describe the relationship between technology, power, and national control.

The review revealed that the current literature is scattered into five distinct groups, but two of them dominate the conversation. One group focuses on the struggle for "digital sovereignty," particularly among countries in the Global South, which are the developing nations often left on the sidelines of the US-China rivalry. The other major group looks at the regulatory competition between the great powers, such as the United States, China, and the European Union, as they try to write their own rules for the future. However, the study found that the issue of how data is extracted and used by powerful companies is often treated as a separate, disconnected problem. Merlet argues that these pieces cannot be understood in isolation. To explain the full picture, the author proposes a new framework called the Geopolitical AI Power Framework. This framework connects three reinforcing forces that work together to create a system where dependency becomes the default outcome.

The first force is structural power asymmetry. This is the idea that power comes from controlling the essential foundations of a system, such as the physical infrastructure, the raw materials, and the rules that govern them. In the world of artificial intelligence, a few actors control the advanced computer chips, the massive cloud computing systems, and the rules for how data moves. Because these resources are so difficult to replicate, countries that do not own them must rely on those who do. This is not just a matter of buying a product; it is about being forced to operate within a structure designed by someone else. The second force is techno-nationalist fragmentation. As countries realize they are dependent, they try to regain control by building their own walls. They pass laws to keep data within their borders, ban foreign platforms, and try to create their own versions of technology. While this feels like a move toward independence, the study suggests it actually makes the problem worse. Instead of one set of global rules, the world is splitting into different blocs, each with its own incompatible standards. Countries without their own technology are then forced to choose a side, further locking them into a specific system.

The third force is the externalities of surveillance capitalism. This term describes a business model where companies turn human behavior and personal data into a product to predict and influence future actions. In the context of artificial intelligence, this means that data flows from less powerful nations to the dominant ones, where it is used to train the very models that those less powerful nations then have to use. This creates a cycle where the dominant actors get smarter and more powerful with every piece of data they receive, while the others become more dependent on the tools built from that data. The study shows that these three forces do not just happen at the same time; they feed into each other. Structural power allows a few nations to set the rules, which causes other nations to fragment into competing blocs, which in turn makes it harder for anyone to escape the data extraction that sustains the original power holders.

The research suggests that this creates a "structural lock-in." This is a tendency, not an unchangeable fate, but it is a powerful one. The more a country adopts foreign artificial intelligence tools without building its own capacity, the harder it becomes to leave that system later. The study finds that while the dominant powers are busy fighting over who leads, the real story is happening in the middle and lower tiers of the global economy. Countries in Latin America, Africa, and Asia are adopting these technologies at high speed, often without the ability to audit or replace them. They are becoming tied to infrastructures and models they did not design, and the decisions made today could fix these dependencies for decades. The review highlights that this is not an accident of history, but a result of how the system is currently designed.

Despite this grim picture, the paper does not argue that the situation is hopeless. The author suggests that semi-peripheral countries—those that are not the superpowers but are not the poorest nations either—still have a window of opportunity. They can shape their position through careful negotiation, by building their own technical capabilities, and by coordinating with their neighbors to create regional solutions. The study points out that some nations are already trying to construct their own paths, showing that dependency is a choice that can be managed, even if it is difficult. However, the window is closing. Every time a government or a company adopts a foreign solution without considering the long-term cost of exit, the chance to build independence shrinks.

The findings of this review serve as a warning and a guide. They suggest that the debate about artificial intelligence needs to move beyond just who has the best chips or the biggest models. The real issue is who controls the conditions under which the technology is used. For a country like Chile, or any nation in a similar position, the question is no longer just about efficiency. It is about who owns the model, where the data lives, and what it would cost to walk away. The study concludes that the global governance of artificial intelligence is fragmenting because the three forces of structural power, regulatory competition, and data extraction are reinforcing one another. Unless there is a deliberate effort to change this cycle, the world is moving toward a future where a few actors set the rules for everyone else, and the rest are left to follow. The research offers a clear map of how we got here, suggesting that understanding these mechanisms is the first step toward ensuring that the future of artificial intelligence does not become a permanent trap for the majority of the world.