The next global power struggle may unfold inside institutions.
GENEVA
Artificial intelligence is changing how institutions exercise authority, although its consequences remain difficult to measure. Governments acquire systems capable of processing information at extraordinary speed, universities incorporate automated tools into research, and corporations increasingly depend on computational models to support decisions. These developments are generally presented as technological progress. Less attention is given to what happens when an institution becomes increasingly dependent on intelligence it cannot independently evaluate.
The difficulty is not confined to countries with limited technological resources. Sophisticated organizations can possess advanced systems while remaining uncertain about the assumptions embedded in their outputs or the conditions under which they might fail. Such uncertainty becomes consequential when technological dependence grows faster than the professional and institutional capacities needed to question it.
A government may expand its computational infrastructure without strengthening its ability to supervise automated decisions. Another institution, operating with fewer resources, could preserve considerable independence through specialized expertise and transparent procedures. Neither arrangement guarantees better governance. The comparison nevertheless challenges conventional assumptions about technological sovereignty.
The Stanford AI Index 2026 documents the concentration of advanced model development within the private sector, particularly among organizations operating in the United States and China. This distribution matters beyond commercial competition. Institutional access to sophisticated artificial intelligence increasingly depends on companies whose technological priorities, commercial incentives and operational decisions may not coincide with those of public authorities.
Yet economic concentration alone does not establish political dependence. Governments have long relied on external suppliers for essential infrastructure without necessarily surrendering institutional authority. Artificial intelligence introduces a different complication: external technologies increasingly participate in producing the information through which institutions interpret problems and evaluate responses.
The geopolitical implications become clearer when examining semiconductors, cloud infrastructure and energy. Advanced chip manufacturing depends on specialized international production networks, including strategically important facilities in Taiwan. Export controls and diplomatic disputes can influence access to essential computational resources. The International Energy Agency also projects that electricity consumption by data centers could approach 945 terawatt-hours by 2030, partly driven by artificial intelligence.
These dependencies reveal a material dimension frequently overlooked in discussions of digital sovereignty. Computational intelligence requires physical infrastructure, financial investment and industrial capabilities distributed unequally across the international system. However, possessing data centers or semiconductor facilities does not automatically establish the institutional capacity to govern the technologies operating through them.
UN Trade and Development projects that the global AI market could reach $4.8 trillion by 2033. Its research also documents substantial concentration in worldwide corporate research and development expenditure, although those figures extend beyond artificial intelligence specifically. The concern is not simply that powerful companies possess significant resources. It is whether their control over critical technologies influences the conditions under which other institutions obtain, evaluate and use them.
This raises a question that technological investment figures cannot resolve. What happens when institutions acquire increasingly sophisticated analytical capabilities without developing equivalent capacities for independent judgment?
Research on human-AI interaction has begun examining automation reliance, cognitive effort and the conditions under which users accept machine-generated recommendations. Some mechanisms designed to encourage critical evaluation can reduce excessive reliance while increasing cognitive workload. Stronger oversight may therefore require additional resources, challenging assumptions that automation necessarily produces both greater efficiency and better decisions.
At the organizational level, the relationship becomes particularly sensitive. Universities may accelerate academic production without establishing whether independent reasoning has improved. Public agencies may generate regulatory analyses more quickly while lacking the expertise to identify unreliable evidence. These possibilities do not demonstrate that artificial intelligence inherently weakens human cognition. They expose the difficulty of distinguishing productivity from intellectual competence.
Such observations suggest a proposed analytical concept: institutional cognitive asymmetry. It describes a potential imbalance in which an organization’s capacity to adopt advanced technological systems expands faster than its capacity to understand, independently evaluate and govern their outputs. The concept is offered as a research proposition requiring empirical validation, not as an established scientific theory.
Its geopolitical relevance emerges when external technological dependence intersects with limited internal expertise. Institutions may become reliant on proprietary models, concentrated suppliers and specialized infrastructure while possessing insufficient knowledge to challenge their operation. Under certain conditions, technological dependence and cognitive asymmetry could reinforce each other, reducing organizational flexibility despite the continued existence of formal authority.
This relationship is neither universal nor inevitable. International cooperation can strengthen institutional capabilities through knowledge transfer, independent training and technological interoperability. Dependence on external providers may be economically rational when credible alternatives and effective oversight remain available. Complete technological self-sufficiency is not necessarily a realistic or desirable measure of sovereignty.
The more consequential distinction concerns informed authority. Can institutions evaluate the systems they acquire, recognize unreliable recommendations and remain operational when technologies become unavailable? Can professionals challenge automated outputs without depending entirely on the same systems for verification? These questions concern organizational competence as much as technological infrastructure.
A comparative research agenda could investigate institutions with similar technological adoption levels but different oversight arrangements. Decision accuracy, professional expertise, supplier concentration, verification procedures and operational resilience would provide relevant dimensions for examination. Longitudinal studies could explore whether AI training strengthens independent judgment or merely increases familiarity with particular platforms.
Causality would require considerable caution. Institutions with weak governance may become technologically dependent because their existing capabilities are limited. Alternatively, prolonged reliance on automation might gradually reduce incentives to maintain internal expertise. Both processes could operate simultaneously, influenced by political conditions, financial resources and organizational culture.
The implications extend beyond emerging economies. Wealthier countries possess greater computational resources but remain exposed to institutional weaknesses, opaque systems and concentrated technological markets. Conversely, organizations operating under financial constraints may preserve meaningful autonomy through professional competence, independent verification and diversified access to technology.
This complicates the geopolitical interpretation of artificial intelligence. Strategic autonomy cannot be reduced to national ownership of infrastructure, nor can technological cooperation automatically be equated with dependence. The capacity to exercise informed judgment may matter as much as the ability to manufacture advanced equipment or finance computational development.
A deeper uncertainty remains. Institutions may become more analytically capable precisely because they delegate increasingly complex tasks to systems they cannot fully explain. Others may preserve extensive human oversight at considerable operational cost. Which arrangement produces greater institutional independence may depend on the decisions involved, the consequences of error and the availability of alternatives.
The emerging geopolitical order is therefore shaped not only by competition over semiconductors, energy and advanced models. It is also influenced by less visible arrangements through which technological systems affect professional knowledge, organizational responsibility and public authority.
Artificial intelligence can strengthen institutional intelligence, but that outcome cannot be inferred from technological adoption alone. Whether external systems expand human judgment or gradually displace essential capacities remains an open empirical question.
Perhaps the most consequential transformation will occur where it is least visible: inside institutions that retain the formal authority to decide, while increasingly depending on systems they cannot independently question.