Managing foreign policy complexity under polycrisis: conceptualising the opportunities and risks of artificial intelligence

Abstract

Contemporary foreign policy is profoundly shaped by a polycrisis, a condition in which political, economic, environmental, technological, and social crises become causally entangled, allowing disturbances in one domain to spread across others and produce greater harm than any single crisis alone. This interconnection exceeds traditional analytical capacities, as foreign ministries are expected to respond faster and across a wider range of entangled issues, often under tight personnel and budgetary constraints. Artificial intelligence has emerged as a promising technology for managing these challenges. Through a conceptual synthesis of interdisciplinary research, this paper argues that the polycrisis has made the capacity to manage complexity an important source of geopolitical advantage, though it has not replaced the older logics of expansion and dominance. Within this framework, AI functions as a central yet ambivalent instrument: while it can strengthen the analytical and operational functions of foreign ministries, it also generates new risks, including a complexity paradox, complexity displacement, crisis-driven entrenchment, and the emergence of an epistemic monoculture. To reconcile these benefits and risks, the paper argues for augmented diplomacy, a model in which AI supplements, rather than replaces, human political judgment.

1 Introduction

The current polycrisis has placed significant pressure on public policies. Foreign policy is among the most affected, as it must mitigate external challenges and formulate resilient responses that safeguard national priorities (Putnam, 1988; Rosenau, 1997; Stanzel, 2018). The interconnection of crisis situations during the current polycrisis has expanded these tasks and made them more complex. Ministries must now respond simultaneously to interconnected challenges, including geopolitical rivalries, intrastate conflicts, financial risks, energy shortages, climate hazards, migratory pressures, and cybersecurity threats (Wiseman, 2019; Heine, 2013; Brosig, 2025; Riordan, 2007). Managing this complexity, rather than addressing concrete threats one at a time, has become a defining challenge of contemporary foreign policy.

At the institutional level, these multiplying demands are testing the capacities of foreign ministries, requiring them to broaden their focus and expertise, adopt faster and more coordinated decision-making, and manage an expanding volume of information. Reforms designed to address these challenges have typically involved organizational restructuring, prioritizing key issue areas, expanding staffing levels, and adopting digital tools (Bátora, 2008; Frissen, 1999; Manor, 2016). These measures are intended to improve responsiveness and help foreign ministries adapt to changing structural conditions. However, while such responses remain context-dependent, ministries operating under different levels of pressure are increasingly expected to accelerate decision-making and preserve professional standards under tightening budgetary constraints (Spence, 1999; Wiseman, 2019; Riordan, 2007).

In this context, recent advances in artificial intelligence (AI) have emerged as a potential instrument for managing complexity and easing pressure on human resources. AI can assist multiple stages of policymaking, from information gathering, analysis, and policy formulation to implementation and evaluation, through faster data collection, automated briefings, trend analysis, cost–benefit calculations, and outcome forecasting (Scott et al., 2018; Chu et al., 2025; Mostafaei et al., 2025). In a field as sensitive as foreign policy, however, the same technology carries significant risks, from flawed or fabricated outputs, automation bias, and limited transparency to the strategic volatility created when many actors depend on similar systems (Atalan et al., 2025; Schmidt, 2022). AI is therefore not only a tool managing complexity but also a source of new complication.

Against this background, this paper aims to provide a conceptual analysis of the role of AI in managing foreign policy complexity under polycrisis. It proceeds through theory synthesis, integrating contributions from geopolitics, foreign policy analysis, institutional theory, digital governance and AI ethics into a unified framework. This framework is centered on the key argument that polycrisis increases the complexity of decision-making environments beyond the capacity of existing institutional and cognitive processes, thereby encouraging the adoption of AI as a corrective tool, while simultaneously introducing new complications (Jaakkola, 2020). Three connected stages develop this argument. The paper first establishes the condition, showing how polycrisis reshapes the geopolitical environment and overloads foreign policy decision-making. It then examines the response, tracing how ministries adapt and how AI is integrated across the stages of the foreign policy process. It finally turns to the complication, assessing the strategic, institutional, and ethical risks that AI generates under polycrisis conditions. Rather than providing an operational roadmap, the paper outlines the conceptual foundations, opportunities, and risks of AI’s role in foreign policy under polycrisis.

2 Research design and methodology

This study adopts a conceptual research design. Rather than testing hypotheses against new empirical evidence, it presents a single argument based on structured reasoning and existing literature. This corresponds with theory synthesis, whereby findings from several fields are integrated to produce an account of a phenomenon that none of them captures on its own (Jaakkola, 2020). In this context, the analysis combines domain theory, which is a specific subject under study, with method theory, which is a broader analytical framework through which that subject is examined. The domain here is the question of polycrisis conditions constrain foreign policy decision-making. The method theory is the growing use of technology, especially AI, as a means of responding to these constraints. These two concepts are examined throughout the paper. The method theory is applied to analyze the phenomenon of domain theory: viewing foreign policy under polycrisis through technological and AI-based responses expands and reorganizes the domain theory, showing which constraints these responses can and cannot ease, and what new constraints they introduce.

The combined analysis, which examines the domain through the lens of technological and AI-based responses, is presented in three stages that trace the phenomenon from its origin to its consequences. As illustrated in Figure 1, the first stage establishes the condition. It examines how the polycrisis reshapes the geopolitical environment, how advantage increasingly depends on adaptation without replacing expansion, and how the resulting complexity burdens foreign policy and its policymaking. The second stage examines the response. It explores how foreign ministries adapt under constraint, in both institutional and individual terms, and how technology can assist this process. It then examines how digital tools, and subsequently AI, emerge as complexity-management instruments across the stages of the foreign policy cycle. The third stage examines the complications. It identifies the risks that AI introduces under polycrisis conditions: the complexity paradox, complexity displacement, crisis-driven entrenchment, and epistemic monoculture. Finally, the paper explores the implications of these findings and presents augmented diplomacy as a model in which AI supports decision-making while human accountability remains central.

Figure 1

3 Polycrisis and the changing geopolitics: challenges for foreign policy

3.1 Polycrisis

The external and internal sphere in which foreign policy operates are increasingly defined by conditions of the polycrisis. The term was introduced by Morin and Kern (1993), who argued that accelerating globalization deepens the interconnection of national systems and infrastructures, allowing local disruptions to create international chain effects and unintended consequences across other areas. These predictions began to materialize in the 2010s, when overlapping geopolitical, economic, environmental, social, digital, and public health crises demonstrated that polycrisis is a multilayered but interconnected condition, in which crises across multiple systems become causally entangled and produce outcomes more harmful than any isolated crisis would (Tooze, 2022; Lawrence et al., 2024). What defines this condition is therefore not the number or frequency of crises themselves, but the relational dynamics between them, as causal interdependence, feedback loops, and mutual amplification allow a disturbance in one domain to spread to others and intensify their effects (Davies and Hobson, 2022; Renn et al., 2019). This relational quality is what separates polycrisis from three adjacent terms of complexity, systemic risk, and the VUCA condition. Complexity refers to the general property of systems built from many interacting parts. Systemic risk refers to the danger that the failure of one component spreads through an entire system. VUCA refers to the volatile, uncertain, complex, and ambiguous conditions under which decisions must be made (Bennett and Lemoine, 2014). Unlike these concepts, polycrisis does not describe a structural property or condition of a system. It refers instead to an active and open-ended situation in which several real crises unfold simultaneously, with no single origin and no foreseeable resolution (Dinan et al., 2024).

Beyond these generalities, the literature is divided on how polycrisis should be conceptualized, what its contents are, and what they mean for humanity and for the decision-makers who need to deal with its consequences. At the broadest level, Søgaard Jørgensen et al. (2024) place the phenomenon in a civilizational frame, identifying a set of self-reinforcing traps that escalating social complexity creates and that limit opportunities of adaptation. At the global level, Lawrence et al. (2024) provide a systematic analysis, concluding that a polycrisis occurs when fast-moving shocks combine with slow-moving stresses and push several global systems out of equilibrium at once. The result is conjoined harm that is greater than the sum of the separate crises, produced by the entanglement of three pathways: shared stresses affecting several systems simultaneously; domino effects carrying disruption from one system to another; and feedback loops through which systems amplify one another. Narrowing the causal question further, Albert (2025) locates the common root of these entangled crises in neoliberal capitalism, using the EU to show how hegemonic constraints limit a polity’s capacity to manage its own crises. Brosig (2025) shifts the focus from structure to agency. While accepting the interconnected nature of polycrisis, the author rejects its extreme interpretations, arguing that crises remain bounded and state leadership continues to shape outcomes. At the narrowest level, sceptics question whether the concept contributes significantly to concrete decision-making, given that the interdependencies that define polycrisis are difficult to identify and act upon (Dinan et al., 2024).

3.2 Changing geopolitics

Although debates on the general nature of the polycrisis remain unresolved, its effects are already reshaping specific domains. Geopolitics is among those that has been affected most directly, along with the policy fields that must respond to it. A useful entry point to comprehend this shift is the classical interpretation of geopolitics, which has been defined by the relationship between power and space. In this interpretation, geography, territory, maritime access, and strategic location determined the relative strengths and vulnerabilities of states (Spykman, 1944; Mackinder, 1904; Mahan, 1890). This classical perspective framed international politics primarily in spatial terms, arguing that geopolitical competition is about the struggle for control over spaces that could improve security, influence, and access to resources (Kelly, 2016). While it treated territory as a mainly objective and fixed reality, critical geopolitics emphasized that space, threat, and national interest are politically constructed through discourse and practice (Koopman et al., 2021; Ó Tuathail, 1996; Ó Tuathail and Dalby, 1998). Rather than focusing only on the distribution of material power, the critical approach examined how political actors interpret and represent the international environment, and how these representations shape threat perceptions, strategic priorities, and the legitimization of specific political actions (Müller, 2008; Dalby, 1991).

These classical and critical approaches demonstrate that geopolitics can be interpreted through multiple lenses, ranging from the distribution of material power to the discursive construction of political action. Nonetheless, these geopolitical theories tend to reflect the historical contexts in which they emerged (Engelbrekt, 2018). Classical geopolitics corresponds to an era of imperial rivalry, industrial expansion, and territorial statecraft, when national greatness was tied to control over land, sea routes, resources, and markets (Nguyen, 2025). Critical geopolitics reflects the late and post-Cold War periods, when growth remained a central objective but became tied to globalization, neoliberal restructuring, and discursive struggles over security, identity, and international order (Ciută and Klinke, 2010; Dalby, 2010). Compared to these historical contexts, contemporary geopolitical dynamics have expanded beyond territorial rivalry and discursive contestation to include various elements of the polycrisis (Scheffran, 2025a, 2025b). Systemic limits have begun to shape geopolitical realities, as parallel and interconnected crisis areas, such as war, migration, pandemics, financial instability, energy insecurity, climate stress, and technological disruption, constrain or exhaust state capacities (Lawrence et al., 2024; Rakowski et al., 2025). As a result, contemporary geopolitical advantage can no longer be understood solely in terms of expansion, control, or dominance. It also depends on the capacity to manage complexity, absorb shocks, stabilize critical systems, reduce dependency, and adapt to a more constrained and volatile international environment (Munoz, 2025; Riordan, 2019; Halpern, 2025).

This does not mean that resilience has displaced the older logic of expansion. The two operate in parallel, as the inherited orientation remains clearly visible in renewed great power competition, military expansion, and the nationalism over critical resources (Nguyen, 2025; Scholvin and Wigell, 2018). However, this expansionist logic now coexists with a second orientation, in which resilience and adaptation become a rational choice. This rationality already drives states to securitize supply chains, diversify national energy resources, and establish technological sovereignty (Edler et al., 2023; Farrell and Newman, 2019). Contemporary geopolitics is therefore transitional, marked by the coexistence of two main orientations. Many states continue to act on the inherited expansionist logic without fully adjusting to the constraints of a finite environment, pursuing the growth-oriented heritage of the 19th and 20th centuries. At the same time, the effects of the polycrisis, such as the COVID-19 or the repeated heatwaves, increasingly rationalize reform, resilience, and sustainability as a form of geopolitical advantage.

3.3 Challenges for foreign policy

This transformation changes not only what states pursue, but also the problems foreign policy must address. While foreign policy has always been a distinct form of public policy, situated at the boundary between domestic and international affairs (Putnam, 1988; Rosenau, 1997), the polycrisis has further eroded the distinction between “inside” and “outside”. By the end of the 20th century, Rosenau (1997) had already argued that globalization, new technologies, and shifting norms were blurring this distinction, strengthening the connection between international and domestic politics. The polycrisis exacerbates this process to the extent where foreign policy issues can no longer be considered as events occurring only in the external environment. Historically, distant international events, such as wars on other continents, migration pressures in neighboring countries, or natural disasters in faraway regions, were treated as external issues, mainly handled by foreign policy institutions, instruments, and experts (Wiseman, 2019; Heine, 2013). Today, these interconnected issues are no longer separable from domestic affairs, as the underlying crises are causally entangled through shared stresses, domino effects, and inter-systemic feedbacks (Lawrence et al., 2024). This generates a ripple effect whereby decisions taken in one field impact those that were once handled independently (Scheffran, 2025b). For instance, a response to an armed conflict can affect energy supply, food security, migration, alliance cohesion, and domestic economic stability simultaneously (Tooze, 2022). The challenge for foreign policy is therefore not only the number of issues to address, but also their growing inseparability, as action on one issue can reshape others in unpredictable ways (Kapucu and Hu, 2022).

These unpredictable interconnections, together with the growing number of issues and tasks, demonstrate how the polycrisis and the consequent shift in geopolitics impact foreign policy through complexity. The resulting strain on the policy machinery is significant. Traditionally, foreign ministries have been organized into regional and functional units, each responsible for its own portfolio and coordinating with the others only when necessary. However, when problems span several domains at once, no single unit owns the problem and coordination across units becomes essential to keep pace with crises. The same pressure affects individual decision-makers, who work under bounded rationality, with limited information, time, and attention, relying on established routines rather than exhaustive analysis (Simon, 1972). The polycrisis amplifies this limitation, as the cross-domain ramifications of any decision surpass what an individual can anticipate. This is where the effects of polycrisis on foreign policy are felt most acutely: while foreign policy struggles to keep the known and the unknown in balance, complexity shifts the workflow toward the latter, pressuring institutions to strengthen both human expertise and technological capabilities in order to anticipate potential outcomes (OECD and World Economic Forum, 2025). The next section is tasked to unfold this problem in an institutional, individual, and technological sense.

4 Adaptation to complexity in foreign affairs

In this complex, crisis-driven geopolitical environment, states bear great responsibility for adopting accurate strategies. Foreign ministries operate under growing complexity described above, where the polycrisis erodes the separability of issues and generates challenges that strains both institutional and individual capacity. Adapting to this pressure is therefore not a secondary administrative task but a condition of remaining effective. The adaptation itself tends to unfold across three connected levels: institutional, individual, and technological (Bátora, 2008; Simon, 1972; Manor, 2016). Drawing on institutional theory, organizational research, public administration, and foreign policy analysis, this section examines the adaptation process at each level.

4.1 Institutional level

At the institutional level, adaptation is shaped by how institutions change over time (Traversa, 2021; Bernhard, 2015). Historical institutionalism emphasizes that institutions develop through historical processes in which earlier decisions tend to shape long-term organizational procedures (Pavanelli and Igari, 2019). Institutional development is therefore path-dependent, as established structures are stabilized by self-reinforcing mechanisms that make change difficult (Farrall, 2021). On the one hand, this resistance hinders fast and structural change if a crisis demands it, particularly when entangled and interconnected issues cannot be easily reorganized around problems that cross every domain at once. On the other hand, path dependency also makes major change more likely at critical turning points, when rising uncertainty and decision-making pressure create opportunities for reforming entrenched practices (Capoccia and Kelemen, 2007). Yet, even in these moments, change is rarely rapid or linear. Decisions made under pressure can shape organizational trajectories for a long time (Beyer, 2026), while reforms may be both contested and promoted by different actors within the institution (Yiangou et al., 2016). As a result, ministries tend to react slowly and unevenly, layering new arrangements onto existing structures rather than replacing them. This creates a persistent gap between foreign ministries’ established procedures and reformed practices, particularly when adaptation requires moving beyond entrenched solutions (Manor, 2016).

Adaptation at this level is also shaped by bureaucratic dynamics within the ministry. According to Gülen (2022), decisions rarely result from a single rational actor; instead, they emerge through bargaining among institutional actors with different mandates and interests. Reorganizing to meet external challenges therefore tends to become a matter of internal contestation, whereby existing units, mandates, or lines of coordination compete with new ones over authority and resources. This produces mixed effects, where rivalry over ownership may lead to fragmentation and duplication, but can also broaden the range of professional perspectives (Hegele, 2018). These dynamics intensify during crises, when information volumes rise and decision-making time contracts. Under such conditions, no single unit can respond alone, and adaptation increasingly depends on whole-of-government coordination across agencies (Christensen et al., 2016; Kapucu and Hu, 2022; Bouckaert and Galego, 2024). In response, many states have reformed their foreign ministries by professionalizing diplomatic services, restructuring institutional frameworks, and incorporating new policy areas such as technology and climate diplomacy (Corsini and Ongaro, 2026). As a result, foreign ministries now operate within broader governmental networks, which create new demands for infrastructure, expertise, and supervision (Hegele, 2018).

4.2 Individual level

At the second, individual level, adaptation is constrained by the cognitive limits of decision-makers. As noted above, officials work under bounded rationality, which holds that decision-makers cannot process all relevant information or evaluate every possible alternative, and therefore rely on simplified rules and heuristics to act under complexity (Simon, 1972; Tsaoussi, 2019; Jonaitytė and Warglien, 2020). These limits become acute during polycrisis conditions, when the volume of information rises, the interdependence among issues deepens, and the time available to decide contracts (Christensen et al., 2016; Zheng, 2025; Moreno, 2021). Officials and the organizations have long sought to manage these pressures through hierarchies, standardized procedures, and collective decision-making, which in turn distribute the cognitive load and impose order on complexity (Ceschi and Fioretti, 2021; Egidi and Sillari, 2020). Yet these established solutions have their limitations when complexity grows. Each additional layer of coordination consumes time and attention, while the recruitment of new staff and the rearrangement of routines deplete ministry resources at a time when available funds are restricted due to the polycrisis. The result is a widening gap between the analysis a situation demands, and the analysis officials can realistically perform. This raises the risk of error, omission, and delay, thus rational decision-makers tend to look beyond established practice, toward instruments that promise to extend the cognitive reach available to them (Bátora, 2008).

4.3 Technological level

At the third, technological level, ministries respond to the limits of institutional and individual adaptation by turning to technological tools. Where restructuring is slow and human cognition is bounded, technology offers a different route, extending the capacity to gather, store, process, and share information beyond what organizations and individuals can manage on their own (Adesina, 2017; Bátora, 2008). In principle, this addresses the problem the two previous levels could not resolve: tools can hold more information in view at once, accelerate its processing, and support coordination across units. Technology therefore appears not as a marginal benefit, but as a tool for bridging the gap between the analysis required by a polycrisis and the resources available to a ministry.

Over the past decades digitalization has offered the most promising technological solutions. Since the 1990s, foreign ministries have introduced digital tools in two distinct ways. They first adapted tools to existing purposes, using new technology to carry out established tasks more efficiently without altering them. They then began to adopt them more deeply, allowing data-driven methods to reshape the tasks themselves and the way diplomatic work is conducted (Bjola and Manor, 2022; Manor, 2019). The concrete results are visible across the range of ministerial tasks. Consular services increasingly rely on chatbots and automated systems to handle routine operations such as visa applications and citizen registration (Pokhriyal and Koebe, 2023). Data analytics and open-source intelligence support policy analysis and offer advantages to states with limited personnel, while geospatial and social-media mapping has helped ministries such as Global Affairs Canada and the United Kingdom’s foreign office locate where their messaging resonates and anticipate emerging consular crises (Kamruzzaman, 2022). In crisis response, satellite imagery and big-data analysis support situational awareness and the coordination of relief, while secure communication systems enable faster exchange across agencies (Manor, 2016). Across these tasks, digitalization has expanded what ministries can do, allowing larger volumes of information to be gathered, processed, and shared than human routines could manage alone.

Yet technological adaptation is subject to the same constraints as the other two levels. Consistent with the path-dependent pattern described above, digital tools have been layered onto existing routines slowly and unevenly, while their benefits have been distributed unequally across states with different resources and infrastructure (Manor, 2016). The introduction of new tools also generated further costs, they required interpretation of additional information, the employment of specialized staff, and integration into slow-to-adapt structures. Consequently, technology supported to ease one constraint while generated another.

With similar benefits and constraints, AI represents the most recent and most powerful step in this trajectory. Where earlier digital tools extended communication and information management, AI is presented as a tool for managing complexity and uncertainty, processing large volume information, modeling alternatives, and supporting decision-making under pressure (OECD and World Economic Forum, 2025; Pokhriyal and Koebe, 2023). This raises the central question of the paper: how can AI support foreign ministries in navigating the polycrisis, where they must address interconnected issues, anticipate cross-domain consequences and make decisions faster than crises unfold? The next section reviews these opportunities by outlining how AI is emerging in, and can be integrated into, foreign affairs, while also examining examples from various countries. Section 6 then turns to the associated limitations and risks.

5 The emergence and integration of AI in foreign policy

5.1 AI as an emergent technology in foreign affairs

Foreign policy has always adapted to new communication technologies, from the telegraph and the telephone to radio and the encrypted cable, each of which changed the speed and reach of diplomatic activity. Digitalization entered this sequence in the late 1990s and has not only accelerated existing practices but restructured the space in which foreign policy is conducted. A significant part of foreign affairs now takes place online, in an environment of platforms, networks, and cloud systems that did not exist a generation ago (Glasze et al., 2023; Gray, 2021). This online environment has become strategically significant where states frame political narratives, extend economic interests, and project cultural influence beyond their borders (Crilley et al., 2020; Manor, 2019; Poell et al., 2025), while also monitoring vulnerabilities, anticipating disruptions, and coordinating responses (Halpern, 2025; Kamruzzaman, 2022).

AI is the most recent technology to enter this environment, offering several different functionalities rather than a single capability (Taeihagh, 2021). Among these, predictive AI uses statistical and machine learning techniques such as regression, classification and clustering to estimate the likelihood of future outcomes. In the field of foreign policy, this may support early warning and risk assessment by anticipating instability, conflict or economic disruption from large volumes of data (Mostafaei et al., 2025). Decision support and discriminative systems use techniques to sort information, detect anomalies and rank options. In policy work, they serve to filter intelligence and structure the alternatives placed before decision-makers (Valle-Cruz et al., 2020). Generative AI, including large language models, probabilistically produces text, imagery, and other content. Although it has mainly been used for general purposes rather than specific tasks (Bommasani et al., 2021), in diplomatic practice it assists with drafting, translation, summarization and discourse analysis.

While these capabilities present both opportunities and risks, and AI solutions depend on offline infrastructure, which has itself become a foreign policy interest. Advanced AI requires large amounts of computing power, advanced semiconductors, data, and energy. The supply chain for these is highly concentrated at every stage: chip design is dominated by a small number of mostly American firms; fabrication is overwhelmingly concentrated in Taiwan; and the necessary lithography technology is supplied by a single European company (Weymouth, 2025). This concentration has turned access to computing power into an issue that ministries must manage diplomatically, through alliances, export control regimes, and tiered licensing that sort states into those granted and those denied access (Edler et al., 2023). The resulting competition directly reflects the shifting geopolitical landscape outlined in Section 3: securing domestic chip production and reducing external dependence demonstrate the logic of resilience, whereas controlling the supply chain and restricting rivals’ access illustrate the strategy of expansion. Pursued either way, these objectives also reflect the interconnected nature of the polycrisis. AI offers opportunities to manage complexity more effectively, yet its offline infrastructure consumes electricity, water, and raw materials, depleting the same finite resources that drive the polycrisis (IEA, 2025; De Vries-Gao, 2026). The emergence of AI therefore generates ambivalent effects: it offers new means of managing complexity while adding to the pressures it is meant to address.

5.2 The integration of AI into the foreign policy cycle

The emergence of AI raises the question of how it can be incorporated into the daily work of foreign policymaking. This integration involves more than introducing new tools to diplomacy. It may also reshape decision-making processes and strategic capabilities across the entire foreign policy cycle, including information gathering and analysis, strategic forecasting, policy formulation, diplomatic communication, implementation, and assessment. This subsection examines each stage in the cycle, considering where AI may contribute and how it might be used. The cycle outlined below should be understood as a heuristic simplification rather than a strictly linear sequence, since the stages actually overlap, recur and vary from one state to another (see Figure 2).

Figure 2

Within the foreign policy cycle, AI may first appear in the information gathering and monitoring, where it can facilitate data collection. In the contemporary foreign policy environment, states and foreign ministries face a growing number of both structured and unstructured data. AI can ease this work. It can process diverse data sources, including social media, online news portals, open-source intelligence (OSINT), and restricted or classified datasets such as cellular data, financial flows, border control records, and satellite observations. These sources generate volumes that traditional tools cannot fully process, while AI-based systems can categorize them rapidly and identify patterns. On this basis, they can perform sentiment analysis, trend detection, and anomaly identification, which may enhance the situational awareness of foreign policy actors (Pokhriyal and Koebe, 2023). These capabilities also support early warning systems play, where AI-based solutions can detect potential shocks earlier than those based on traditional data processing tools. Established projects such as the Integrated Crisis Early Warning System (ICEWS) and the Political Instability Task Force have used machine learning to predict armed conflicts, while the Rhombus system signaled the likelihood of Russia’s invasion of Ukraine months in advance (De Agostini and Giovanardi, 2025). Although a correct signal alone does not establish general predictive capacities, these examples suggest that AI may narrow the gap between detection and response.

While gathering information is crucial, it is meaningless without accurate interpretation. In the policymaking process, raw data must be interpreted, contextualized and transformed into strategic decisions. Generative and decision-support systems may assist here, as they can draft reports, summarize complex documents, and identify patterns in raw data, thus providing information needed to make decisions (Mostafaei et al., 2025). Large volumes of data can be examined for trends or correlations that human analyst might find difficult to consider. Consequently, foreign ministries may treat AI as a support tool, particularly in forecasting, strategic planning, and negotiation preparation (Mamakou et al., 2025). However, this analytical role has limits. Foreign policy decisions are not purely technical optimization problems, as many have no single correct answer and are shaped by the normative and strategic interests of those involved (Jensen et al., 2025). AI may therefore support analytical processes, but the interpretation of its outputs remains human responsibility.

Interpreting present information is still not sufficient, since advantage under polycrisis may also depend on anticipating developments that cannot yet be observed. This is the basis of the next stage, strategic forecasting and scenario planning. Foreign policymaking in such conditions tends to follow an anticipatory logic, in which states gain advantage by foreseeing escalation or geopolitical shifts and preparing for them. Predictive systems may support this by generating probabilistic forecasts and simulating different geopolitical scenarios from available data, which could assist conflict prevention and preventive diplomacy by identifying potential risks and outcomes before they materialize (De Agostini and Giovanardi, 2025). These forecasts are not neutral, however, Jensen et al. (2025) developed benchmark systems to examine the foreign policy preferences and escalation tendencies of different AI models. Their research demonstrates that AI models may produce different responses to identical geopolitical scenarios. This indicates that AI systems are not impartial instruments but may contain implicit preferences and biases embedded in their training data and model architecture. The forecasting stage therefore illustrates both potential contribution of AI and the caution its outputs require.

Data gathering, interpretation, and forecasting may contribute directly to policy formulation and decision support. At this stage, AI can help to structure decision-making by generating and comparing policy options, aligning alternatives with national and political interests, conducting cost–benefit and impact analyses, and assessing the likely outcomes of different scenarios (Scott et al., 2018). Its role here is not to determine the final decision, but to organize the available choices in a way that makes them comparable and reduces the cognitive burden on decision-makers during complex deliberations. Decision-support systems are most directly relevant at this stage, as they sort and rank options rather than generate content or forecasts. In practice, this requires a division of labor: AI may structure the deliberative process, while human actors retain responsibility for the ultimate political and normative decision that follows (Manor, 2026).

Once a course of action has been chosen, it must be carried out and communicated, and AI can support both. In terms of implementation, AI may assist real-time coordination among agencies, facilitate information sharing, and help allocate resources during the execution of a decision, which is matters most in crisis response, where speed and coordination are essential (Kapucu and Hu, 2022). Communication is the second aspect of this stage, since few foreign policy decisions take effect without being conveyed to other states, institutions, or publics. The rise of digital diplomacy has been significant here, as the digital sphere is now where narratives are shaped and influence is exercised. Generative systems can support these tasks by integrating multilingual communication, targeted audience engagement, and optimized digital campaigns, while also enabling public diplomacy to analyze public attitudes, online discourse, and opinion trends (Mamakou et al., 2025). However, these same capabilities carry clear risks: generative models increase the potential for manipulation through deepfakes and disinformation, meaning the communication aspect of implementation has become algorithm-driven, both in production and reception (Brundage et al., 2018).

The cycle closes with assessment and repositioning, where the results of a decision are evaluated and integrated into future policy. AI may support this stage by monitoring the implementation of a decision, measuring outcomes against stated objectives, and identifying discrepancies between results and expectations (Valle-Cruz et al., 2020; OECD, 2025). The analytical and pattern-recognition capacities used at earlier stages can be applied here to evaluation, providing a more systematic basis for determining whether a policy has achieved its goals (Whitsel et al., 2024). The value of this stage lies in the feedback it provides, as the findings of assessment return to the information gathering and analysis stage and allows states to adjust their strategies as conditions change. This feedback also makes the integration of AI a continuous process rather than a single intervention (Valle-Cruz et al., 2020). Such continuity may matter under polycrisis, where the rapid pace of change requires constant revision of strategies instead of a one-time implementation (Søgaard Jørgensen et al., 2024).

The recurring principle across all six stages is that AI may support foreign policy decision-making, but full automation without human involvement is neither likely nor desirable. Despite technological change and digitalization, foreign policy decisions still require cultural, historical, and political judgment, simple data processing is not sufficient on its own (Vera Hoyos and Cárdenas Marín, 2025). The simplified integration described here therefore indicates a division of labor, in which AI produces analyses, generates options, and provides forecasts, while human actors interpret these outputs, validate them, and take responsibility for the decisions (Guru, 2025). This pairing allows ministries to utilize algorithmic capacity while retaining control over strategic and normative choices (Manor, 2026). The following examples illustrate how this integration has begun to take shape in practice, before the analysis turns to the risks it carries.

5.3 Evidence from state practices

In this section, four illustrative examples are utilized to demonstrate different forms of AI implementation in foreign affairs. Singapore, Estonia, the United Kingdom, and the United States have each integrated AI into governance and diplomacy in distinct ways, and together they show the range of institutional, infrastructural, and operational forms this integration can take. The case studies were not selected as a representative sample, but as illustrative examples, each of which highlights a different precondition or dimension of AI-supported foreign policy.

Singapore’s model is characterized by strong state coordination, extensive public–private collaboration, and carefully designed governance institutions. Rather than focusing exclusively on technological innovation, the government has invested substantial resources in building the institutional capacities necessary for the responsible deployment of AI. These include regulatory frameworks, human capital development, trustworthy data infrastructures, and cross-sector partnerships. As a result, AI is not treated merely as a technological tool but as a strategic governance resource that strengthens national decision-making while maintaining political accountability (Goode et al., 2023; Cheng et al., 2026). The same approach is reflected in Singapore’s foreign policy. Rather than allowing AI to independently determine foreign policy priorities, Singapore primarily employs it to support strategic foresight, information processing, policy coordination, and long-term national resilience. Cheng et al. (2026) argue that Singapore’s AI strategy is inseparable from the country’s middle-power foreign policy objectives, particularly its ambition to preserve strategic autonomy amid the technological competition between the United States and China. The 2C1D framework (Collaboration—Contribution—Diversification), developed by the authors, demonstrates that the development of AI capabilities is closely intertwined with the expansion of diplomatic relations, international partnerships, and multilateral cooperation, rather than being driven by a pursuit of technological self-sufficiency. Sear (2026) further reinforces this argument by emphasizing the role of institutional builders, professionals capable of designing, developing, and strengthening the institutions needed to govern emerging technologies effectively. This implies that effective AI governance relies on human oversight, but several skills are essential for decision-makers including the necessary professional expertise, adequate AI literacy, appropriate institutional authority, and the genuine ability to critically assess, modify, or even reject algorithm-generated recommendations. These skills ensure that human oversight to be a substantive mechanism.

Estonia’s example illustrates the importance of institutional preparedness in AI-driven diplomacy. This preparedness rests on a robust and comprehensive digital state architecture that enables AI-supported governance, secure data exchange, and cross-border digital cooperation. A cornerstone of this model is the X-Road infrastructure, which enables secure, time-stamped, and interoperable data exchange among public institutions. According to Hardy (2024), the international diffusion of Estonia’s X-Road system, particularly through cooperation with Finland, Iceland, the Faroe Islands, and the Åland Islands, can itself be regarded as a form of digital diplomacy. The small Baltic state also illustrates the infrastructural dimension of technological integration. Its digital identity system, the X-Road platform, the Data Embassy initiative, and a range of cross-border interoperability projects provide the secure, interoperable, and traceable foundations needed for integrating AI into public administration and decision-support processes. As a result, AI can strengthen governmental capacity without replacing political decision-making. Another illustrative example is the Bürokratt initiative, which aims to develop a nationwide AI-powered digital assistant that connects citizens with public services through an interoperable network of government systems (Thian, 2025). Dreyling et al. (2024) argue that the project demonstrates the importance of organizational agility, open collaboration, and citizen-centered design in the successful implementation of AI within the public sector. At the same time, the Eesti.ai initiative seeks to increase national productivity through the wider adoption of AI, with particular emphasis on education, healthcare, and security (e-Estonia, 2026). Particularly noteworthy is its proposal to establish digital identities for AI agents, with the aim of making them uniquely identifiable and traceable (Henning, 2026). Such an approach makes it possible to determine which AI system was responsible for a specific decision or action, thereby establishing clear lines of responsibility and accountability. Estonia’s case points out that the human-in-the-loop element is not enough and should be accompanied by a sophisticated institutional arrangement that ensure the transparency, traceability, and accountability of AI systems throughout their operation.

The United Kingdom provides an example of how AI can be integrated into diplomatic practice, including diplomatic service delivery, crisis forecasting, consular services, climate security, and the shaping of international AI governance and regulatory frameworks. In his 2025 speech, “Diplomacy in the Digital Age,” David Lammy, the UK’s Foreign Secretary, emphasized that diplomacy must increasingly operate in an information environment characterized by “machine speed” (Hemrajani and Tan, 2025). At the same time, he stressed that the human dimension of diplomacy must be preserved under all circumstances, underscoring that AI should enhance rather than replace human judgment, trust, and political decision-making. At the operational level, the Foreign, Commonwealth and Development Office (FCDO) have already deployed a range of AI-enabled solutions to enhance its diplomatic and consular services. One of the best-known examples is the AI-driven Correspondence Triage system, which automatically classifies, prioritizes, and routes incoming correspondence (Cavanaugh, 2024). Following its implementation, processing times were reportedly reduced from approximately ten days to just a matter of seconds. This example illustrates the administrative and analytical dimensions of AI-supported diplomacy. Rather than generating diplomat responses autonomously, AI accelerates the organization, filtering, and processing of available information, enabling diplomats to devote more time to tasks that require professional judgment, political sensitivity, negotiation skills, and strategic decision-making (Department for Science, Innovation and Technology, 2026). Beyond administrative efficiency, the UK also employs AI to strengthen strategic foresight and crisis preparedness. The government has launched an international AI partnership aimed at improving global preparedness for climate-change related shocks (Hunt, 2026). The initiative includes data sharing, the development of AI-based predictive models, capacity building, and technical support for national meteorological services. In this context, AI becomes a diplomatic instrument that enhances resilience, facilitates the early identification of emerging risks, and supports international cooperation in addressing shared global challenges (Truby et al., 2026). A UK policy report focusing on the Middle East and North Africa (MENA) further reinforces this perspective. According to the report, the UK intends to leverage AI as a tool of foreign policy influence in areas including energy optimization, health technologies, workforce development, and AI governance (Cambridge Middle East and North Africa Forum, 2025). The report conceptualizes AI as a fundamental instrument of statecraft, arguing that the UK can offer partner countries an AI governance model founded on transparency, trust, and democratic values. From this perspective, AI is understood not only as a technological capability but as a strategic foreign policy asset that strengthens diplomatic influence while promoting rules-based international cooperation and responsible digital governance.

The United States is widely regarded as one of the most advanced practical examples of AI-supported diplomacy. AI has become increasingly embedded in the day-to-day operations of the U.S. Department of State, yet its purpose is not to replace diplomats but to enhance their effectiveness. In this context, AI functions as a force multiplier that significantly improves information processing, analytical capacity, strategic foresight, and administrative efficiency, while foreign policy decisions remain firmly under human control (Vota, 2024). The institutional foundation is provided by the U.S. Department of State Enterprise Artificial Intelligence Strategy (U.S. Department of State, 2023). According to the strategy, the primary objective of AI is to strengthen diplomacy through its responsible use (Kilburg, 2025). To achieve this, AI is intended to support a wide range of functions, including knowledge management, data analytics, multilingual communication, information processing, and evidence-based decision support (Kruchoski, 2026). Consequently, AI is conceptualized not as an autonomous decision-maker but as an organizational capability that enhances the work of diplomats. According to the American Foreign Service Association (AFSA), U.S. diplomats now routinely employ generative AI to support numerous everyday tasks (Kruchoski, 2026). These include drafting diplomatic cables, summarizing policy documents, preparing country reports, processing political analyses, supporting research activities, producing communication materials, and multilingual translation. The adoption of generative AI has substantially reduced the time devoted to information-intensive administrative work, allowing diplomats to focus more extensively on core diplomatic responsibilities such as negotiations, relationship-building, and strategic political analysis (Zia, 2025). Similarly, the USC Center on Public Diplomacy argues that AI increasingly supports a broad range of diplomatic functions, including predictive analytics, scenario simulation, negotiation preparation, sentiment analysis, and real-time information processing (Vota, 2024). Therefore, the role of diplomats is not diminished but, rather, becomes even more significant.

These cases reveal several common features in how foreign ministries integrate and work with AI. Singapore demonstrates that the successful integration of AI into foreign policy depends less on technological sophistication than on strong institutional capacity and governance. Estonia highlights the importance of secure digital infrastructure, traceability, and accountability mechanisms as prerequisites for meaningful human oversight. The United Kingdom illustrates how AI can enhance the speed, strategic foresight, and service capacity of diplomacy while preserving human judgment and institutional accountability. Finally, the United States exemplifies a functional division of labor in which AI performs information-intensive analytical tasks, whereas diplomats retain responsibility for political judgment, negotiations, ethical deliberation, and final decision-making. Combined, these cases support the argument for division of labor identified in the previous subsection. They indicate that the integration of AI strengthened the analytical and operational capabilities of diplomatic institutions, yet strategic authority, political responsibility, and democratic accountability remained firmly under human control.

6 Risks, limits, and ethical challenges of AI in foreign policy

The integration described in the previous section also carries significant challenges. The first part of this section reviews the established risk categories that are widely discussed in the public policy literature and are also relevant to foreign policy. The second, based on the conceptual framework developed earlier in this paper, identifies the further risks that arise specifically when AI is adopted under polycrisis conditions.

6.1 General risks and limitations of AI

The most basic limitations are epistemic and operational. Contemporary machine learning models tend to generate outputs without offering transparent explanations of the analytical processes through which they were produced (Chu et al., 2025; Scott et al., 2018). This lack of transparency is not a temporary technical limitation, but rather a structural weakness of many algorithmic systems (Burrell, 2016; Zednik, 2021). Closely related is the bias inherited from training data: since AI systems learn from existing records, their outputs reflect the analytical frameworks and structural inequalities embedded in those datasets (Alon-Barkat and Busuioc, 2023; Bender et al., 2021). This risk is particularly acute in foreign policy. Diplomatic and intelligence datasets often overrepresent the perspectives of powerful states and are organized through analytical traditions developed in the Global North, leaving alternative strategic cultures underrepresented (Okolo and Raji, 2026; Roberts et al., 2021; Tironi and Albornoz, 2025). A further operational concern is reliability. Generative models can produce fluent yet inaccurate outputs that become hazardous in time-pressured policy environments, where verification routines are among the first activities to be cut (Pokhriyal and Koebe, 2023; Bender et al., 2021).

These limits lead directly to problems of accountability and governance. If AI systems become embedded in workflows as routine analytical tools, operators may begin to trust algorithmic outputs even when other sources suggest otherwise. This automation bias gradually weakens the practical distinction between advisory input and human decision (Horowitz and Kahn, 2024; Alon-Barkat and Busuioc, 2023; Parasuraman and Manzey, 2010) and risks a slow erosion of analytical capacity within foreign ministries, as routine judgment is delegated to automated systems (Pokhriyal and Koebe, 2023). The resulting accountability problem worsens when responsibility for an outcome is distributed across institutional hierarchies that use different levels of AI modeling, thereby producing errors that are difficult to identify. Current legal and institutional frameworks offer little guidance on who is responsible when an AI-assisted assessment contributes to a policy failure, leaving a gap that requires legislative attention (Floridi and Cowls, 2019; Chesterman, 2021).

Beyond these internal limits, there is a structural asymmetry in access to and control over AI capabilities. As noted earlier, advanced models, computational infrastructures, and human expertise remain concentrated in a limited number of states and private corporations (Brundage et al., 2018; Dafoe, 2018). This concentration produces unequal positions among states, where technologically capable governments can process intelligence more quickly, detect hostile behavior more accurately, and deploy responses that less capable states may find difficult to counter. A parallel asymmetry separates states and the private sector. Most foreign ministries depend on commercial providers for their AI tools, and core analytical capacities are therefore hosted by entities whose interests, accountability, and data governance practices may not align with legal and public norms (Scott et al., 2018; Binns, 2018).

Reliance on a narrow set of providers also creates risks related to dependence and predictability. It can make ministerial decision-making vulnerable to supply disruptions, changes to service terms, and unilateral vendor modifications. A related hazard concerns the behavior of AI systems in high-stake scenarios. Wargame simulations using large language models have shown tendencies toward escalation and unpredictable patterns of arms-race behavior, indicating that models adapted for foreign policy applications may behave more aggressively than human decision-makers (Lamparth et al., 2024; Rivera et al., 2024). Predictability poses a further risk: if adversaries can anticipate how a model will respond to certain inputs, they can manipulate those inputs to gain advantage over conventional intelligence operations (Brundage et al., 2018).

Underlying all of these is a set of ethical tensions concerning legitimacy, fairness, and the role of human judgment. The opacity of AI systems, combined with commercial confidentiality, creates structural barriers to legislative oversight, public deliberation, and judicial review, especially when ministries use tools whose internal logic cannot be disclosed publicly (Deeks, 2025; Couldry and Mejias, 2019; Floridi and Cowls, 2019). Fairness concerns arise when AI-influenced decisions discriminate, either positively or negatively, against foreign societies, for example, in the allocation of development aid, the prioritization of humanitarian assistance, the targeting of sanctions, and the assessment of migration risk (Crawford, 2021; Eubanks, 2018). International initiatives such as the 2021 UNESCO Recommendation on the Ethics of AI and the 2023 Bletchley Declaration recognize these tensions, yet the gap between nonbinding principles and enforceable international regulations remains significant (Taeihagh, 2021; Cihon et al., 2020).

While these five categories provide a useful overview, they primarily describe risks that AI poses across public policy, including foreign policy. Building on the conceptual framework developed earlier in this paper, the rest of this section identifies a further set of risks that arise specifically when AI is adopted in under polycrisis conditions (see Figure 3).

Figure 3

6.2 Emergent risks under polycrisis

The first is the complexity paradox. Foreign ministries are adopting AI precisely because the volume, speed, and complexity of contemporary policy demands have surpassed human analytical capacity. Yet AI does not only mitigate complexity; it also generates new layers that ministries must subsequently manage. These include regulatory complexity (compliance, audit, and liability frameworks), infrastructural complexity (data storage, computing capacity, energy supply, and cooling), human resource complexity (AI requires technical staff whose professional cultures differ from traditional diplomatic profiles), and supply chain complexity (most ministries depend on a small number of foreign suppliers for essential hardware and models) (Bommasani et al., 2021). The tool adopted to reduce complexity therefore becomes a generator of further complications, partially redistributing the analytical burden of foreign ministries from the core work of foreign policy to the management of supporting technology.

Complexity displacement extends this problem from the institutional to the systemic level. As Section 3 argued, contemporary geopolitical advantage depends on the simultaneous management of overlapping crises across political, economic, environmental, and social sectors. AI affects several of these domains directly. Depending on scenario, data centers are projected to consume 2–4% of global electricity production by 2035 (IEA, 2025). The accompanying demand for water, raw materials, and grid capacity collides with environmental sustainability efforts, especially when digital expansion relies on carbon-intensive sources (De Vries-Gao, 2026; Bender et al., 2021). Labor market disruption associated with AI-driven automation may likewise generate social and political tensions that governments must address (Kauhanen and Rouvinen, 2025; Lane and Saint-Martin, 2021). Consequently, using a single technology to manage the complexity of polycrisis may help address certain challenges while simultaneously intensifying pressures in other areas, such as energy systems, environmental sustainability, and labor markets. AI is therefore not only a complexity-management but also a complexity-displacement instrument, with the displaced complexity tending to surface elsewhere within the same polycrisis.

Building on the historical institutionalist framework discussed earlier, a third risk, crisis-driven entrenchment, emerges from the timing of AI adoption. Crises produce turning points at which institutional choices become difficult to reverse, and decisions taken under such conditions determine the institutional trajectory for extended periods (Capoccia and Kelemen, 2007; Pierson, 2000). AI is currently being integrated into foreign ministries under exactly such crisis-driven conditions, where ministerial capacity is stretched and reform is adaptive rather than deliberate. The danger is not only that ministries adopt AI poorly, but that they adopt it now, under the least favorable conditions, locking themselves into operational strategies, model architectures, and corporate arrangements that will outlive the present crisis. This entrenchment is significant, as switching from one foundation model or cloud provider to another requires substantial costs, data conversion, staff retraining, and workflow disruption (Bommasani et al., 2021). The adoption of high-consequence, low-reversibility infrastructure under crisis conditions is therefore institutionally hazardous.

The final risk concerns the emergence of an epistemic monoculture across foreign ministries. This danger is associated with the loss of analytical diversity: if ministries across multiple states rely on a limited number of models supplied by a few providers, two ministries asking similar questions of similar models will tend to receive similar answers (Bommasani et al., 2021). Research on algorithmic monoculture highlights this issue, demonstrating that when multiple decision-makers rely on the same algorithm or underlying models, their decisions homogenize, making the system prone to correlated failures, even when the shared model is recommending the accurate response (Kleinberg and Raghavan, 2021; Bommasani et al., 2022). Consequently, there is a risk that diversity declines and analytical errors become correlated. This is a particularly concerning risk during a polycrisis, when challenges spread through causally interdependent systems. Historically, states have responded to these dynamics as distributed actors, interpreting the international environment through their own traditions and detecting signals that others overlook (Page, 2008; Holling, 1973). A shared analytical foundation eliminates this dispersion: a blind spot, bias or error in the common models leads to a collective failure across multiple foreign policies, which manifest simultaneously rather than independently. In this case, AI no longer mitigates complexity but instead becomes one of its main causes. The shared models form a new transmission channel within the polycrisis through which a single analytical failure can spread across states (Lawrence et al., 2024). Therefore, the technology adopted to manage the entanglement of crises may become a source of that entanglement, effectively turning AI into a mechanism of the polycrisis that it was intended to contain.

7 Implications: an argument for augmented diplomacy

The findings of this paper identify several challenges associated with the introduction of AI into contemporary foreign policy. These include the complexity generated by polycrisis, the institutional and individual limitations of managing it, the integration of AI across the policy cycle, and the risks that follow from this process. When combined, these separate findings converge on a single conclusion: AI can extend the analytical and operational capacity of foreign ministries, but it cannot assume the contextual, normative, and political judgment on which foreign policy decisions depend, and wherever it is allowed to do so, new risks follow. This convergence supports a particular model of division of labor between humans and technology, which earlier research has labeled augmented diplomacy (Manor, 2026). The concept refers to hybrid arrangements in which AI extends human reach by taking over information-intensive tasks that have become unmanageable due to their complexity, while retaining human control over judgment, authority, and responsibility. Its premise is that the two contributions are complementary rather than interchangeable (Picavet et al., 2026). AI provides processing scale and speed: it performs information-intensive and analytical tasks such as document summarization, multilingual processing and translation, knowledge management, pattern recognition, predictive analytics, scenario modeling, negotiation preparation and sentiment analysis. Human actors exercise judgment: they retain exclusive responsibility for interpreting political context, conducting negotiations, defining strategic priorities, exercising ethical consideration and assuming final political accountability (Guru, 2025; Vera Hoyos and Cárdenas Marín, 2025). Nevertheless, the effectiveness of this division depends on whether human oversight is substantive rather than formal. Diplomats should not simply approve AI-generated recommendations, but must critically evaluate the underlying evidence, assumptions, potential biases and alternative interpretations before determining a course of action (Laux and Ruschemeier, 2025; Alon-Barkat and Busuioc, 2023).

That said, these interpretations require competence, and competence requires ministries either to hire AI experts or to train diplomats. In both cases, AI literacy is needed to understand what these systems can and cannot do, regional and political expertise to contextualize their outputs, legal and ethical awareness to judge compliance with international and national law, institutional authority to question, modify, or reject recommendations, and sufficient time and capacity to assess them independently (Hemrajani and Tan, 2025). Given this complexity, oversight by individuals is not enough on its own. Accountability also depends on institutional arrangements that ensure transparency, traceability, and clearly defined responsibility across the AI-workflow, including audit trails, explicit chains of responsibility, and mechanisms such as the digital identities for AI agents proposed in Estonia (Henning, 2026). A human-in-the-loop approach should therefore be complemented by institutional governance that distributes accountability across diplomats, technical units, and the private providers on which ministries increasingly depend.

The effectiveness of augmented diplomacy can be assessed through measurable indicators, including decision-processing time, forecasting accuracy, the rate at which AI recommendations are modified or rejected, bias detection, the transparency and traceability of AI-supported decisions, and the degree of institutional learning over time (Floridi and Cowls, 2019; Valle-Cruz et al., 2020; Laux and Ruschemeier, 2025). The United Kingdom’s AI-enabled Correspondence Triage system, which reduced processing times from days to seconds while remaining under human supervision, illustrates how such improvements can be both real and measurable without displacing human control (Cavanaugh, 2024). Considering this example, augmented diplomacy can be viewed as a promising solution for ministries, enabling them to leverage algorithmic capacity to manage complexity while retaining authority over strategic and normative choices (Manor, 2026).

The previous discussion outlined a diverse set of tasks and ministerial functions that concern not only diplomats and classical diplomacy, but also the wider foreign policy apparatus, including desk officers, analysts and bureaucrats. This raises questions about the scope of the concept. Although the term augmented diplomacy places the diplomat at its center, most of the examined functions, such as information gathering, analysis, forecasting, drafting and assessment, are performed within the ministry rather than by envoys in the field. The clearest examples, ranging from the United Kingdom’s correspondence triage to the United States’ internal knowledge management, relate to the administrative and analytical core of foreign ministries rather than diplomatic representation abroad. The terminology of diplomacy has always incorporated this broader institutional apparatus, and the literature on digital diplomacy reflects this interpretation (Manor, 2016; Adesina, 2017). The findings of this paper reinforce the expansive approach and indicate that AI is extending across the entire ministerial workflow and apparatus, not just the work of envoys. Augmented diplomacy should therefore be considered to cover the broad field of foreign affairs rather than diplomatic representation in its narrow sense.

8 Conclusion

This paper has offered a conceptual analysis of how foreign policy can manage complexity under polycrisis through AI. The argument developed in three connected steps. It first established the condition, showing how the polycrisis reshapes the geopolitical environment and generates a complexity that foreign ministries are increasingly unable to manage through existing routines. It then examined the response, tracing how ministries adapt institutionally, individually, and technologically, and how AI is integrated across the stages of the foreign policy cycle. It finally turned to the complication, assessing the risks that this integration generates in a sensitive policy area where decisions carry strategic, normative, and international legal weight.

The analysis indicated a changing nature of geopolitical advantage. While classical and critical perspectives interpreted geopolitics through power, space, and the discursive construction of interests, the polycrisis redefines advantage as the capacity to manage complexity, absorb shocks, and stabilize critical systems. Foreign ministries face these requirements with institutions designed for separate issues and the bounded attention of individual decision-makers, which naturally leads them towards digital technology and, most recently, AI.

The central finding is that integrating AI into foreign policy can ease complexity at several stages of the policy process, but it also generates new risks. Beyond the established limitations of AI, the conditions of polycrisis produce additional and distinctive risks. AI may exert additional pressure in areas such as energy and labor, constrain ministries into low-reversibility decisions, and, as states converge on a limited number of shared models, erode the analytical diversity upon which collective resilience depends. In each case, the technology adopted to manage complexity also reproduces it, which is the central tension the paper has sought to expose. As argued in the previous section, augmented diplomacy offers a way to balance these benefits and risks, allowing ministries to utilize AI’s capabilities while preserving human judgment.

These conclusions emerge from conceptual rather than empirical analysis. This is a real limitation, together with the paper’s reliance on illustrative rather than representative cases, its inability to test specific AI systems across ministerial contexts, and its lack of access to ministerial practices closed to public study. While these limitations constrain the scope of the argument, they do not undermine it. The mechanisms on which the paper relies, namely bounded rationality, path dependence, automation bias, and algorithmic monoculture, have each been studied, and the contribution here is to show how they interact in the specific context of foreign policy under polycrisis. Building on these foundations, future research should test these conceptual principles against practice, examining how ministries in different states work with these systems and, where possible, reaching the institutional practices that this study could not observe. This matters because AI is already being integrated into foreign policy, and it is advancing faster than the field has been able to study it. In this context, the paper demonstrated that AI can no longer be considered an external instrument designed to ease complexity, because it has become part of the complexity itself. AI now both relieves and adds to the pressures on foreign ministries, and distinguishing between these two effects becomes the central task. This is why human judgment and interpretation remain central, not only in practice but also in academic research, which helps practitioners understand where AI genuinely supports foreign policy and where it introduces new risks.

References

• 1

  AdesinaO. S. (2017). Foreign policy in an era of digital diplomacy. Cogent Soc. Sci.3:1297175. doi: 10.1080/23311886.2017.1297175

  • CrossRef
  • Google Scholar

• 2

  AlbertM. J. (2025). Capitalism, complexity, and polycrisis: toward neo-Gramscian polycrisis analysis. Glob. Sustain.8:e7. doi: 10.1017/sus.2025.10

  • CrossRef
  • Google Scholar

• 3

  Alon-BarkatS.BusuiocM. (2023). Human–AI interactions in public sector decision making: ‘automation bias’ and ‘selective adherence’ to algorithmic advice. J. Public Adm. Res. Theory33, 153–169. doi: 10.1093/jopart/muac007

  • CrossRef
  • Google Scholar

• 4

  AtalanY.ReynoldsI.JensenB. (2025). AI Biases in Critical Foreign Policy Decisions. Washington, DC: Center for Strategic and International Studies.

  • Google Scholar

• 5

  BátoraJ. (2008). Foreign Ministries and the Information Revolution: Going Virtual? Leiden: Brill Nijhoff. Available online at: https://brill.com/display/title/14859 (Accessed May 30, 2026).

  • Google Scholar

• 6

  BenderE. M.GebruT.McMillan-MajorA.ShmitchellS. (2021). On the Dangers of Stochastic Parrots: Can Language Models be too big?Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–623.

  • Google Scholar

• 7

  BennettN.LemoineG. J. (2014). What a difference a word makes: understanding threats to performance in a VUCA world. Bus. Horiz.57, 311–317. doi: 10.1016/j.bushor.2014.01.001

  • CrossRef
  • Google Scholar

• 8

  BernhardM. (2015). Chronic instability and the limits of path dependence. Perspect. Politics13, 976–991. doi: 10.1017/S1537592715002261

  • CrossRef
  • Google Scholar

• 9

  BeyerJ. (2026). On a branching route: the spectrum of path dependence in policy research. Rev. Policy Res.43:e70007. doi: 10.1111/ropr.70007

  • CrossRef
  • Google Scholar

• 10

  BinnsR. (2018). Fairness in machine learning: lessons from political philosophy. Proc. Mach. Learn. Res.81149–159. Available online at: https://proceedings.mlr.press/v81/binns18a.html

  • Google Scholar

• 11

  BjolaC.ManorI. (2022). The rise of hybrid diplomacy: from digital adaptation to digital adoption. Int. Aff.98, 471–491. doi: 10.1093/ia/iiac005

  • CrossRef
  • Google Scholar

• 12

  BommasaniR.CreelK. A.KumarA.JurafskyD.LiangP. (2022). Picking on the same person: does algorithmic monoculture lead to outcome homogenization?Adv. Neural Inf. Proces. Syst.35, 3663–3678. doi: 10.48550/arXiv.2211.13972

  • CrossRef
  • Google Scholar

• 13

  BommasaniR.HudsonD. A.AdeliE.AltmanR.AroraS.von ArxS.et al. (2021). On the opportunities and risks of foundation models. arXiv, 1–214. Available online at: https://arxiv.org/pdf/2108.07258

  • Google Scholar

• 14

  BouckaertG.GalegoD. (2024). System-quake proof ‘systemic resilience governance’: six measures for readiness. Glob. Policy15, 97–105. doi: 10.1111/1758-5899.13433

  • CrossRef
  • Google Scholar

• 15

  BrosigM. (2025). From neologism to promising research agenda? The global polycrisis and IR. Int. Relat. [Preprint]. 00471178251333294. doi: 10.1177/00471178251333294

  • CrossRef
  • Google Scholar

• 16

  BrundageM.AvinS.ClarkJ.TonerH.EckersleyP.GarfinkelB.et al. (2018). The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation. Cambridge: Apollo—University of Cambridge Repository.

  • Google Scholar

• 17

  BurrellJ. (2016). How the machine ‘thinks’: understanding opacity in machine learning algorithms. Big Data Soc.3:2053951715622512. doi: 10.1177/2053951715622512

  • CrossRef
  • Google Scholar

• 18

  Cambridge Middle East and North Africa Forum (2025). UK AI Diplomacy: Boosting British Influence in MENA and Strengthening the Abraham Accords. Cambridge: Cambridge Middle East and North Africa Forum and UK Abraham Accords Group.

  • Google Scholar

• 19

  CapocciaG.KelemenR. D. (2007). The study of critical junctures: theory, narrative, and counterfactuals in historical institutionalism. World Polit.59, 341–369. doi: 10.1017/S0043887100020852

  • CrossRef
  • Google Scholar

• 20

  CavanaughL. (2024). UK bets big on AI-driven Diplomatic Services, November 19. GovInsider. Available online at: https://govinsider.asia/intl-en/article/uk-bets-big-on-ai-driven-diplomatic-services (Accessed June 30, 2026)

  • Google Scholar

• 21

  CeschiA.FiorettiG. (2021). From bounded rationality to collective behavior. Nonlinear Dyn. Psychol. Life Sci.25, 385–394. Available online at: https://www.societyforchaostheory.org/ndpls/askFILE/?docObjId=250401ABSTRACT

  • Google Scholar

• 22

  ChengJ. H.ChongB.GomezM. A. (2026). Strategic choices for middle powers in developing AI capabilities: a case study of Singapore. Asian Secur.22, 16–32. doi: 10.1080/14799855.2025.2581269

  • CrossRef
  • Google Scholar

• 23

  ChestermanS. (2021). We, the Robots? Regulating Artificial Intelligence and the Limits of the Law. 1st Edn. Cambridge: Cambridge University Press.

  • Google Scholar

• 24

  ChristensenT.LægreidP.RykkjaL. H. (2016). Organizing for crisis management: building governance capacity and legitimacy. Public Adm. Rev.76, 887–897. doi: 10.1111/puar.12635

  • CrossRef
  • Google Scholar

• 25

  ChuK.ElangoK.SpillaneM.ChanS.SiganA.NadlerN.et al. (2025). AI and the Future of Foreign PolicyInstitute for Youth in Policy. Available online at: https://yipinstitute.org/policy/ai-and-the-future-of-foreign-policy (Accessed May 30, 2026).

  • Google Scholar

• 26

  CihonP.MaasM. M.KempL. (2020). Should artificial Intelligence Governance be Centralised?Design Lessons from History Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society228–234.

  • Google Scholar

• 27

  CiutăF.KlinkeI. (2010). Lost in conceptualization: reading the ‘new cold war’ with critical geopolitics. Polit. Geogr.29, 323–332. doi: 10.1016/j.polgeo.2010.06.005

  • CrossRef
  • Google Scholar

• 28

  CorsiniH.OngaroE. (2026). Bureaucratic responses to populist government: explaining foreign policy (non-)change. JCMS J. Common Mark. Stud.64, 24–47. doi: 10.1111/jcms.13748

  • CrossRef
  • Google Scholar

• 29

  CouldryN.MejiasU. A. (2019). Making data colonialism liveable: how might data’s social order be regulated?Internet Policy Rev.8. doi: 10.14763/2019.2.1411

  • CrossRef
  • Google Scholar

• 30

  CrawfordK. (2021). The Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. New Haven, CT: Yale University Press.

  • Google Scholar

• 31

  CrilleyR.ManorI.BjolaC. (2020). Visual narratives of global politics in the digital age: an introduction. Camb. Rev. Int. Aff.33, 628–637. doi: 10.1080/09557571.2020.1813465

  • CrossRef
  • Google Scholar

• 32

  DafoeA. (2018). AI Governance: A Research Agenda. Oxford: Governance of AI Program, Future of Humanity Institute, University of Oxford.

  • Google Scholar

• 33

  DalbyS. (1991). Critical geopolitics: discourse, difference, and dissent. Environ. Plan. D Soc. Space9, 261–283. doi: 10.1068/d090261

  • CrossRef
  • Google Scholar

• 34

  DalbyS. (2010). Recontextualising violence, power and nature: the next twenty years of critical geopolitics?Polit. Geogr.29, 280–288. doi: 10.1016/j.polgeo.2010.01.004

  • CrossRef
  • Google Scholar

• 35

  DaviesM.HobsonC. (2022). An embarrassment of changes: international relations and the COVID-19 pandemic. Aust. J. Int. Aff.77, 150–168. doi: 10.1080/10357718.2022.2161990

  • CrossRef
  • Google Scholar

• 36

  De AgostiniL.GiovanardiM. (2025). AI and Global Security: From Early Warning to AI-Assisted Diplomacy. Task Force 4: Global Peace and Security Policy Brief. Think7. Available online at: https://www.think7.org/publications/ai-and-global-security-from-early-warning-to-ai-assisted-diplomacy/ (Accessed May 30, 2026).

  • Google Scholar

• 37

  De Vries-GaoA. (2026). The carbon and water footprints of data centers and what this could mean for artificial intelligence. Patterns7:101430. doi: 10.1016/j.patter.2025.101430,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 38

  DeeksA. S. (2025). The Double Black Box: National Security, Artificial Intelligence, and the Struggle for Democratic Accountability. 1st Edn New York, NY: Oxford University Press.

  • Google Scholar

• 39

  Department for Science, Innovation and Technology (2026). AI Opportunities Action Plan: One Year On. Policy Paper. UK Government. Available online at: https://www.gov.uk/government/publications/ai-opportunities-action-plan-one-year-on/ai-opportunities-action-plan-one-year-on (Accessed June 30, 2026)

  • Google Scholar

• 40

  DinanS.BélandD.HowlettM. (2024). How useful is the concept of polycrisis? Lessons from the development of the Canada emergency response benefit during the COVID-19 pandemic. Policy Design Pract.7, 430–441. doi: 10.1080/25741292.2024.2316409

  • CrossRef
  • Google Scholar

• 41

  DreylingR.TammetT.PappelI.McBrideK. (2024). Navigating the AI maze: lessons from Estonia's Bürokratt on public sector AI digital transformation. SSRN Electron. J. [Preprint]. doi: 10.2139/ssrn.4850696

  • CrossRef
  • Google Scholar

• 42

  EdlerJ.BlindK.KrollH.SchubertT. (2023). Technology sovereignty as an emerging frame for innovation policy: defining rationales, ends and means. Res. Policy52:104765. doi: 10.1016/j.respol.2023.104765

  • CrossRef
  • Google Scholar

• 43

  e-Estonia. (2026). Estonia Launches bold AI Initiative – Eesti.ai. Available online at: https://e-estonia.com/estonia-ai-initiative-launch-eesti-ai (Accessed June 30, 2026).

  • Google Scholar

• 44

  EgidiM.SillariG. (2020). “Bounded rationality and organizational decision making,” in Routledge Handbook of Bounded Rationality, ed. VialeR. (London: Routledge), 509–521.

  • Google Scholar

• 45

  EngelbrektK. (2018). A brief intellectual history of geopolitical thought and its relevance to the Baltic Sea region. Glob. Aff.4, 475–485. doi: 10.1080/23340460.2018.1535256

  • CrossRef
  • Google Scholar

• 46

  EubanksV. (2018). Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor. New York, NY: Macmillan.

  • Google Scholar

• 47

  FarrallS. (2021). “Historical and constructivist institutionalisms,” in Building Complex Temporal Explanations of Crime, ed. FarrallS. (Cham: Palgrave Macmillan), 29–50.

  • Google Scholar

• 48

  FarrellH.NewmanA. L. (2019). Weaponized interdependence: how global economic networks shape state coercion. Int. Secur.44, 42–79. doi: 10.1162/isec_a_00351

  • CrossRef
  • Google Scholar

• 49

  FloridiL.CowlsJ. (2019). A unified framework of five principles for AI in society. Harvard Data Sci. Rev.1. doi: 10.1162/99608f92.8cd550d1

  • CrossRef
  • Google Scholar

• 50

  FrissenP. H. A. (1999). Politics, Governance and Technology: A Postmodern Narrative on the Virtual State. Cheltenham: Edward Elgar Publishing.

  • Google Scholar

• 51

  GlaszeG.CattaruzzaA.DouzetF.DammannF.BertranM.-G.BômontC.et al. (2023). Contested spatialities of digital sovereignty. Geopolitics28, 919–958. doi: 10.1080/14650045.2022.2050070

  • CrossRef
  • Google Scholar

• 52

  GoodeK.KimH. M.DengM. (2023). Examining Singapore's AI Progress (Issue Brief). Center for Security and Emerging Technology (CSET). Available online at: https://cset.georgetown.edu/publication/examining-singapores-ai-progress (Accessed June 30, 2026)

  • Google Scholar

• 53

  GrayJ. E. (2021). The geopolitics of ‘platforms’: the TikTok challenge. Internet Policy Rev.10. doi: 10.14763/2021.2.1557

  • CrossRef
  • Google Scholar

• 54

  GülenB. (2022). Turf wars in foreign policy bureaucracy: rivalry between the government and the bureaucracy in Turkish foreign policy. Foreign Policy Anal.18:orac021. doi: 10.1093/fpa/orac021

  • CrossRef
  • Google Scholar

• 55

  GuruA. (2025). The Future of Diplomacy: AI’S Expanding role in International Affairs Observer Research Foundation. Available online at: https://www.orfonline.org/expert-speak/the-future-of-diplomacy-ai-s-expanding-role-in-international-affairs (Accessed May 30, 2026).

  • Google Scholar

• 56

  HalpernO. (2025). The geo-politics of resilience: on the historical convergence between ecology, artificial intelligence, and corporate strategy. New Media Soc.27, 4581–4605. doi: 10.1177/14614448251336420,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 57

  HardyA. (2024). Estonia’s digital diplomacy: Nordic interoperability and the challenges of cross-border e-governance. Int. Policy Rev.13. doi: 10.14763/2024.3.1785

  • CrossRef
  • Google Scholar

• 58

  HegeleY. (2018). Explaining bureaucratic power in intergovernmental relations: a network approach. Public Adm.96, 753–768. doi: 10.1111/padm.12537

  • CrossRef
  • Google Scholar

• 59

  HeineJ. (2013). “From club to network diplomacy,” in The Oxford Handbook of Modern Diplomacy, eds. CooperA.HeineJ.ThakurR. (Oxford: Oxford University Press).

  • Google Scholar

• 60

  HemrajaniA.TanR. (2025). The Role of AI in Modern Diplomacy. RSIS Commentary No. 199. S. Rajaratnam School of International Studies (RSIS), Nanyang Technological University. Available online at: https://rsis.edu.sg/wp-content/uploads/2025/09/CO25199.pdf (Accessed June 30, 2026).

  • Google Scholar

• 61

  HenningM. (2026). Estonia to give digital Identities to AI Agents. Euractiv. Available online at: https://www.euractiv.com/news/estonia-to-give-digital-identities-to-ai-agents (Accessed June 30, 2026).

  • Google Scholar

• 62

  HollingC. S. (1973). Resilience and stability of ecological systems. Annu. Rev. Ecol. Syst.4, 1–23. doi: 10.1146/annurev.es.04.110173.000245

  • CrossRef
  • Google Scholar

• 63

  HorowitzM. C.KahnL. (2024). Bending the automation bias curve: a study of human and AI-based decision making in national security contexts. Int. Stud. Q.68:sqae020. doi: 10.1093/isq/sqae020

  • CrossRef
  • Google Scholar

• 64

  HuntM. (2026). UK Government to use AI to Improve Global Preparation for Climate Shocks. Global Government Forum. Available online at: https://www.globalgovernmentforum.com/uk-government-to-use-ai-to-improve-global-preparation-for-climate-shocks (Accessed June 30, 2026).

  • Google Scholar

• 65

  IEA (2025). Energy and AI. Paris: International Energy Agency.

  • Google Scholar

• 66

  JaakkolaE. (2020). Designing conceptual articles: four approaches. AMS Rev.10, 18–26. doi: 10.1007/s13162-020-00161-0

  • CrossRef
  • Google Scholar

• 67

  JensenB.ReynoldsI.AtalanY.GarciaM.WooA.ChenA.et al. (2025). Critical foreign policy decisions (CFPD)-benchmark: measuring diplomatic preferences in large language models. arXiv, 1–50. Available online at: https://arxiv.org/pdf/2503.06263

  • Google Scholar

• 68

  JonaitytėI.WarglienM. (2020). “Attention and organizations,” in Routledge Handbook of Bounded Rationality, ed. VialeR. (London: Routledge), 522–534.

  • Google Scholar

• 69

  KamruzzamanM. M. (2022). Impact of social media on geopolitics and economic growth: mitigating the risks by developing artificial intelligence and cognitive computing tools. Comput. Intell. Neurosci.2022, 1–12. doi: 10.1155/2022/7988894,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 70

  KapucuN.HuQ. (2022). An old puzzle and unprecedented challenges: coordination in response to the COVID-19 pandemic in the US. Public Perform. Manag. Rev.45, 773–798. doi: 10.1080/15309576.2022.2040039

  • CrossRef
  • Google Scholar

• 71

  KauhanenA.RouvinenP. (2025). Assessing early labour market effects of generative AI: evidence from population data. Appl. Econ. Lett., 1–4. doi: 10.1080/13504851.2025.2513973

  • CrossRef
  • Google Scholar

• 72

  KellyP. (2016). Classical Geopolitics: A New Analytical Model. Stanford, CA: Stanford University Press.

  • Google Scholar

• 73

  KilburgD. (2025). AI Use Cases for Diplomats: Applying Artificial Intelligence to Diplomacy. 1st Edn. Boca Raton, FL: Chapman & Hall/CRC Press.

  • Google Scholar

• 74

  KleinbergJ.RaghavanM. (2021). Algorithmic monoculture and social welfare. Proc. Natl. Acad. Sci.118:e2018340118. doi: 10.1073/pnas.2018340118,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 75

  KoopmanS.DalbyS.MegoranN.SharpJ.KearnsG.SquireR.et al. (2021). Critical geopolitics/critical geopolitics 25 years on. Polit. Geogr.90:102421. doi: 10.1016/j.polgeo.2021.102421

  • CrossRef
  • Google Scholar

• 76

  KruchoskiP. (2026). Leading from the edge: how diplomats are actually using AI. Foreign Serv. J. Available online at: https://afsa.org/leading-edge-how-diplomats-are-actually-using-a

  • Google Scholar

• 77

  LamparthM.CorsoA.GanzJ.MastroO. S.SchneiderJ.TrinkunasH. (2024). Human vs. machine: behavioral differences between expert humans and language models in wargame simulations. Proc. AAAI/ACM Conf. AI Ethics Soc. 7, 807–817. doi: 10.1609/aies.v7i1.31681

  • CrossRef
  • Google Scholar

• 78

  LaneM.Saint-MartinA. (2021). The Impact of Artificial Intelligence on the Labour Market: What Do We Know So Far? OECD Social, Employment and Migration Working Papers No. 256. Paris: OECD Publishing.

  • Google Scholar

• 79

  LauxJ.RuschemeierH. (2025). Automation bias in the AI act: on the legal implications of attempting to de-bias human oversight of AI. Europ. J. Risk Regulation16, 1519–1534. doi: 10.1017/err.2025.10033,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 80

  LawrenceM.Homer-DixonT.JanzwoodS.RockströmJ.RennO.DongesJ. F. (2024). Global polycrisis: the causal mechanisms of crisis entanglement. Glob. Sustain.7:e6. doi: 10.1017/sus.2024.1,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 81

  MackinderH. J. (1904). The geographical pivot of history. Geogr. J.23:421. doi: 10.2307/1775498

  • CrossRef
  • Google Scholar

• 82

  MahanA. T. (1890). The Influence of Sea Power upon History, 1660–1783. Boston, MA: Little, Brown and Company.

  • Google Scholar

• 83

  MamakouI. M.MarkantoniA.KargasA. D. (2025) Leveraging AI in Diplomacy: Enhancing Foreign Policy, national Interest and digital Diplomacy. 33rd European Regional ITS Conference, Edinburgh, UK

  • Google Scholar

• 84

  ManorI. (2016). Are We There Yet: Have MFAs Realized the Potential of Digital Diplomacy? Results from a Cross-National Comparison. Brill Research Perspectives. Leiden: Brill.

  • Google Scholar

• 85

  ManorI. (2019). The Digitalization of Public Diplomacy. Cham: Palgrave Macmillan.

  • Google Scholar

• 86

  ManorI. (2026). From digital diplomacy to AI-driven diplomacies? Mapping the potential impact of AI on diplomacy. Commun. Public. doi: 10.1177/20570473261438571

  • CrossRef
  • Google Scholar

• 87

  MorenoO. M. C. (2021). Coordinated governance in the VUCA scenario. Hrvat. Komparat. Javna Uprava21, 393–422. doi: 10.31297/HKJU.21.3.6

  • CrossRef
  • Google Scholar

• 88

  MorinE.KernA. B. (1993). Terre-Patrie. Paris: Éditions du Seuil.

  • Google Scholar

• 89

  MostafaeiH.KordnooriS.OstadrahimiM.BanihashemiS. S. A. (2025). Applications of artificial intelligence in global diplomacy: a review of research and practical models. Sustain. Fut.9:100486. doi: 10.1016/j.sftr.2025.100486

  • CrossRef
  • Google Scholar

• 90

  MüllerM. (2008). Reconsidering the concept of discourse for the field of critical geopolitics: towards discourse as language and practice. Polit. Geogr.27, 322–338. doi: 10.1016/j.polgeo.2007.12.003

  • CrossRef
  • Google Scholar

• 91

  MunozJ. M. (2025). The Geopolitics and Geoeconomics of Technology. 1st Edn. London: Routledge.

  • Google Scholar

• 92

  NguyenH. P. G. (2025). The international turn in the study of the origins of late-19th-century imperialism: achievements, limits and new directions in the theory of uneven and combined development. Int. Relat. [Preprint]:00471178251319705. doi: 10.1177/00471178251319705

  • CrossRef
  • Google Scholar

• 93

  Ó TuathailG. (1996). Critical Geopolitics: The Politics of Writing Global Space, vol. 6London: Routledge.

  • Google Scholar

• 94

  Ó TuathailG.DalbyS. (1998). “Introduction: rethinking geopolitics: towards a critical geopolitics,” in Rethinking Geopolitics, eds. Ó TuathailG.DalbyS. (London: Routledge).

  • Google Scholar

• 95

  OECD (2025). Competition in Artificial Intelligence Infrastructure. OECD Competition Law and Policy Working Papers. Paris: OECD Publishing.

  • Google Scholar

• 96

  OECD and World Economic Forum (2025). AI in Strategic Foresight: Reshaping Anticipatory Governance. Paris: OECD Publishing.

  • Google Scholar

• 97

  OkoloC. T.RajiM. (2026). “The global majority in international AI governance,” in Handbook on the Global Governance of Artificial Intelligence, eds. FurendalM.LundgrenM. (Cheltenham: Edward Elgar Publishing), 133–153. doi: 10.4337/9781035338580.00016

  • CrossRef
  • Google Scholar

• 98

  PageS. (2008). The Difference: How the Power of Diversity Creates Better Groups, Firms, Schools, and Societies. Princeton, NJ: Princeton University Press.

  • Google Scholar

• 99

  ParasuramanR.ManzeyD. H. (2010). Complacency and bias in human use of automation: an attentional integration. Hum. Factors52, 381–410. doi: 10.1177/0018720810376055,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 100

  PavanelliJ. M. M.IgariA. T. (2019). Institutional reproduction and change: an analytical framework for Brazilian electricity generation choices. Int. J. Energy Econ. Policy9, 252–263. doi: 10.32479/ijeep.8056

  • CrossRef
  • Google Scholar

• 101

  PicavetM. E. B.MaroniP.SandhuA.DesouzaK. C. (2026). Human–machine collaboration for strategy foresight: the case of generative AI. Public Adm. Rev.86, 299–310. doi: 10.1111/puar.70048

  • CrossRef
  • Google Scholar

• 102

  PiersonP. (2000). Increasing returns, path dependence, and the study of politics. Am. Polit. Sci. Rev.94, 251–267. doi: 10.2307/2586011

  • CrossRef
  • Google Scholar

• 103

  PoellT.DuffyB. E.NieborgD. B.MutsvairoB.TseT.ArriagadaA.et al. (2025). Global perspectives on platforms and cultural production. Int. J. Cult. Stud.28, 3–20. doi: 10.1177/13678779241292736

  • CrossRef
  • Google Scholar

• 104

  PokhriyalN.KoebeT. (2023). AI-assisted diplomatic decision-making during crises: challenges and opportunities. Front. Big Data6:1183313. doi: 10.3389/fdata.2023.1183313,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 105

  PutnamR. D. (1988). Diplomacy and domestic politics: the logic of two-level games. Int. Organ.42, 427–460. doi: 10.1017/S0020818300027697

  • CrossRef
  • Google Scholar

• 106

  RakowskiJ. J.SchaanL. N.Van KlinkR.HerzonI.ArthA.HagedornG.et al. (2025). Characterizing the global polycrisis: a systematic review of recent literature. Annu. Rev. Environ. Resour.50, 159–183. doi: 10.1146/annurev-environ-111523-102238

  • CrossRef
  • Google Scholar

• 107

  RennO.LucasK.HaasA.JaegerC. (2019). Things are different today: the challenge of global systemic risks. J. Risk Res.22, 401–415. doi: 10.1080/13669877.2017.1409252

  • CrossRef
  • Google Scholar

• 108

  RiordanS. (2007). Reforming foreign services for the twenty-first century. Hague J. Dip.2, 161–173. doi: 10.1163/187119007X180539

  • CrossRef
  • Google Scholar

• 109

  RiordanS. (2019). The Geopolitics of Cyberspace: A Diplomatic Perspective. Leiden: Brill.

  • Google Scholar

• 110

  RiveraJ.-P.MukobiG.ReuelA.LamparthM.SmithC.SchneiderJ. (2024). Escalation Risks from Language Models in Military and Diplomatic Decision-MakingProceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency836–898.

  • Google Scholar

• 111

  RobertsH.CowlsJ.MorleyJ.TaddeoM.WangV.FloridiL. (2021). The Chinese approach to artificial intelligence: an analysis of policy, ethics, and regulation. AI Soc.36, 59–77. doi: 10.1007/s00146-020-00992-2

  • CrossRef
  • Google Scholar

• 112

  RosenauJ. N. (1997). Along the Domestic-Foreign Frontier: Exploring Governance in a Turbulent World. Cambridge: Cambridge University Press.

  • Google Scholar

• 113

  ScheffranJ. (2025a). “Planetary boundaries, polycrisis and politics in the Anthropocene: climate pathways, tipping cascades and transition to sustainable peace in integrative geography,” in Towards Rethinking Politics, Policy and Polity in the Anthropocene, ed. BrauchH. G., vol. 35 (Cham: Springer Nature Switzerland).

  • Google Scholar

• 114

  ScheffranJ. (2025b). Systemic risks and governance of the global polycrisis in the Anthropocene: stability of the climate–conflict–migration–pandemic nexus. Glob. Sustain.8:e39. doi: 10.1017/sus.2025.10026

  • CrossRef
  • Google Scholar

• 115

  SchmidtN. (2022). Artificial Intelligence and Foreign Policy Challenges. Prague: Centre for Governance of Emerging Technologies, Institute of International Relations.

  • Google Scholar

• 116

  ScholvinS.WigellM. (2018). Power politics by economic means: geoeconomics as an analytical approach and foreign policy practice. Comp. Strateg.37, 73–84. doi: 10.1080/01495933.2018.1419729

  • CrossRef
  • Google Scholar

• 117

  ScottB.HeumannS.LorenzP. (2018). Artificial Intelligence and Foreign PolicyBerlinStiftung Neue Verantwortung—Beisheim Center. Available online at: https://www.interface-eu.org/storage/archive/files/ai_foreign_policy.pdf (Accessed May 30, 2026).

  • Google Scholar

• 118

  SearM. J. (2026). Why the AI-driven Future Requires Institutional Builders, not Technologists. GovInsider. Available online at: https://govinsider.asia/intl-en/article/why-the-ai-driven-future-requires-institutional-builders-not-technologists (Accessed June 30, 2026).

  • Google Scholar

• 119

  SimonH. A. (1972). Theories of bounded rationality. Dec. Org.1, 161–176. Available online at: https://www.scirp.org/reference/referencespapers?referenceid=2129523

  • Google Scholar

• 120

  Søgaard JørgensenP.JansenR. E. V.Avila OrtegaD. I.Wang-ErlandssonL.DongesJ. F.ÖsterblomH.et al. (2024). Evolution of the polycrisis: Anthropocene traps that challenge global sustainability. Philos. Trans. R. Soc. B379:20220261. doi: 10.1098/rstb.2022.0261,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 121

  SpenceD. (1999). “Foreign ministries in national and European context,” in Foreign Ministries: Change and Adaptation, ed. HockingB. (London: Palgrave Macmillan), 247–268.

  • Google Scholar

• 122

  SpykmanN. J. (1944). The Geography of the Peace. New York, NY: Harcourt, Brace and Company.

  • Google Scholar

• 123

  StanzelV. (2018). New Realities in Foreign Affairs: Diplomacy in the 21st Century. Berlin: German Institute for International and Security Affairs.

  • Google Scholar

• 124

  TaeihaghA. (2021). Governance of artificial intelligence. Polic. Soc.40, 137–157. doi: 10.1080/14494035.2021.1928377

  • CrossRef
  • Google Scholar

• 125

  ThianS. Y. (2025). Estonia eyes cross-Border Interoperability for Bürokratt, its 'Siri of public Services'. GovInsider. Available online at: https://govinsider.asia/intl-en/article/estonia-eyes-cross-border-interoperability-for-burokratt-its-siri-of-public-services (Accessed June 30, 2026).

  • Google Scholar

• 126

  TironiM.AlbornozC. (2025). Decolonizing AI? Lessons from a failed experiment. Big Data Soc.12:20539517251365224. doi: 10.1177/20539517251365224

  • CrossRef
  • Google Scholar

• 127

  ToozeA. (2022). Welcome to the World of the Polycrisis. Financial Times. Available online at: https://www.ft.com/content/498398e7-11b1-494b-9cd3-6d669dc3de33 (Accessed July 21, 2026).

  • Google Scholar

• 128

  TraversaF. (2021). Power and institutional change: from path dependence to theories of gradual change. Rev. Econ. Inst.23, 83–108. doi: 10.18601/01245996.v23n45.05

  • CrossRef
  • Google Scholar

• 129

  TrubyJ.DahdalA.BrownR.IbrahimI. (2026). Diplomacy in the age of AI: legal and strategic approaches to techno-nationalism, regulatory soft power and the AI chips race. Res. Glob.12:100335. doi: 10.1016/j.resglo.2026.100335

  • CrossRef
  • Google Scholar

• 130

  TsaoussiA. (2019). “Bounded rationality,” in Encyclopedia of Law and Economics, ed. BackhausJ. (New York, NY: Springer), 147–151.

  • Google Scholar

• 131

  U.S. Department of State (2023). Enterprise Artificial Intelligence Strategy FY2024–FY2025: Empowering Diplomacy through Responsible AI U.S. Department of State. Available online at: https://www.state.gov/wp-content/uploads/2023/11/Department-of-State-Enterprise-Artificial-Intelligence-Strategy.pdf (Accessed June 30, 2026).

  • Google Scholar

• 132

  Valle-CruzD.CriadoJ. I.Sandoval-AlmazánR.Ruvalcaba-GomezE. A. (2020). Assessing the public policy-cycle framework in the age of artificial intelligence: from agenda-setting to policy evaluation. Gov. Inf. Q.37:101509. doi: 10.1016/j.giq.2020.101509

  • CrossRef
  • Google Scholar

• 133

  Vera HoyosC.Cárdenas MarínW. O. (2025). The use of artificial intelligence in political decision-making. Philosophies10:95. doi: 10.3390/philosophies10050095

  • CrossRef
  • Google Scholar

• 134

  VotaW. (2024). How US State Department Uses AI Strategically in Modern Diplomacy. ICTworks. Available online at: https://www.ictworks.org/state-department-uses-ai-strategically (Accessed June 30, 2026).

  • Google Scholar

• 135

  WeymouthS. (2025). Digital disintegration: techno-blocs and strategic sovereignty in the AI era. Int. Organ.79, S57–S70. doi: 10.1017/S0020818325101070

  • CrossRef
  • Google Scholar

• 136

  WhitselL. P.HoneycuttS.RadcliffeR.JohnsonJ.ChaseP. J.NoyesP. (2024). Policy implementation and outcome evaluation: establishing a framework and expanding capacity for advocacy organizations to assess the impact of their work in public policy. Health Res. Policy Syst.22:27. doi: 10.1186/s12961-024-01110-0,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 137

  WisemanG. (2019). Contemporary challenges for foreign ministries: at home and abroad. Diplom. Statecraft30, 786–798. doi: 10.1080/09592296.2019.1673554

  • CrossRef
  • Google Scholar

• 138

  YiangouJ.O’KeeffeM.GlöcklerG. (2016). “‘Tough love’: how the ECB’S monetary financing prohibition pushes deeper euro area integration,” in Redefining European Economic Governance, eds. ChangM.MenzG.SmithM. P. (London: Routledge).

  • Google Scholar

• 139

  ZednikC. (2021). Solving the black box problem: a normative framework for explainable artificial intelligence. Philos. Technol.34, 265–288. doi: 10.1007/s13347-019-00382-7

  • CrossRef
  • Google Scholar

• 140

  ZhengH. (2025). Crisis-driven governance reforms: an analytical framework of institutional capacity, leadership, and collaboration in global governance. Chin. Public Adm. Rev.17, 19–38. doi: 10.1177/15396754251335231

  • CrossRef
  • Google Scholar

• 141

  ZiaL. (2025). Negotiating with Algorithms: The Future of AI-Powered Diplomacy. USC Center on Public Diplomacy, University of Southern California. Available online at: https://uscpublicdiplomacy.org/blog/negotiating-algorithms-future-ai-powered-diplomacy (Accessed June 30, 2026).

  • Google Scholar