AI, Geopolitics And Framework

Overview

Policymakers are striving to understand the use, associated risks and benefits of AI technologies. The technologies are developing at pace and often involve highly technical details. Coumpounded by the persistent hype around AI, this complexity can make it challenging for non-specialists to understand AI’s impact, including in a geopolitical context. Yet, the need to respond to disruptive technological developments is not new: International Relations scholarship has often provided useful tools, frameworks or concepts to understand how such changes impact interstate relations. Can policymakers turn to such scholarship to provide valid analytical tools to better understand AI’s impact on geopolitics? This workshop report summarises the discussion of 10 scholars, debating what impact AI has on geopolitics, whether there is a need for a new analytical framework to capture AI’s impact on international relations, and what we can learn from other technologies and their respective impact on geopolitical developments.

Introduction

In November 2023, RUSI hosted an invitation-only workshop to discuss whether there is a genuine need for a new analytical framework in International Relations scholarship to grasp AI’s impact on geopolitics. RUSI convened a deliberately small number of leading International Relations scholars working on AI and other technologies to discuss the impact of AI on international relations, what can be learned from previous technological inventions on the interplay of technology and international relations, and whether a new framework or concept to analyse AI’s impact on geopolitics is needed.

The workshop included 10 scholars and was chaired by Anthony Finkelstein, President of City, University of London, and a RUSI Distinguished Fellow. In addition to the findings from the workshop itself, invited participants were asked to fill out a survey in advance of the workshop (please see the Annex for the survey questions). Of the 10 participants who attended the workshop, six provided a survey response. An additional two responses were provided by those who were invited but did not attend. This workshop report relies on the contributions from both the survey and the workshop. All views expressed in the workshop and the survey are non-attributable.

Context

AI technologies, their risks and opportunities, as well as their regulation and ethical challenges, are highly debated topics in UK policy circles. This is particularly the case since the launch of the UK National AI Strategy, setting out the aim of making the UK an “AI superpower”, and in the context of the UK AI Safety Summit held in November 2023. Policymakers in the UK increasingly turn to questions of AI benefits, implications and risks, including its relationship with foreign affairs and international relations.

Too often, however, it is difficult for policymakers and non-technicians to grasp AI’s novel implications for familiar fields like International Relations. This is partially due to the knowledge gap on how AI technology works and how it will develop. This gap exists between technology experts and those in the social sciences and policy circles.

Are the tools and methods currently available adequate to frame and explain AI’s impact on international relations? Or are new ones required to capture AI’s implications for interstate relations? Only where there are adequate methods to understand and communicate such impact will non-technicians and policymakers be appropriately equipped to grasp and communicate the implications that follow from AI technologies for international relations. These might be considered decisively different or similar to those of existing technologies. Reaching such understanding is, of course, a necessary pre-condition for policymakers to design effective policy interventions for AI technologies that will allow countries such as the UK to secure a strategic advantage from AI by harnessing its advantages and mitigating any arising risks.

Academic scholarship can provide suitable and widely applied frameworks and concepts to understand technologies’ impact on international relations – such as, for example, deterrence theory as it applies to nuclear weapons and their impact on international relations during the Cold War. Even where such theory’s ability to accurately capture a technology’s impact on international relations is debatable, it can nevertheless serve as a starting point of discussion.

AI technologies are not a completely new topic in social sciences and International Relations. That said, the recent trend has certainly led to more academic attention to AI. Whereas much academic discussion has traditionally focused on ethical implications of so-called “killer robots” and the use of AI by lethal autonomous weapons systems (LAWS), scholars in the social sciences are now expanding their research to include areas such as AI governance and AI-enabled disinformation campaigns.

RUSI’s workshop included International Relations scholars working on the implications of AI, or other technologies such as space or nuclear technologies, for international relations. Participants were invited to discuss three subject areas:

1. AI’s impact on geopolitics (Part One).

2. What can be learned from the way International Relations theory and scholarship has explained previous technological inventions (Part Two).

3. Whether a new framework is needed to conceptualise AI’s impact on international relations (Part Three).

Part One: AI’s Impact on Geopolitics and International Relations

There was widespread agreement that AI has a noticeable and manifold impact on international relations as well as International Relations scholarship – not least because states are addressing issues of AI governance and risk management. The impacts have also been felt as AI potentially has an effect on interstate power dynamics.

Examples of how AI impacts geopolitical dynamics included:

• Applications in the military domain, including strategic stability, the laws and ethics of war, and the public perception of warfare.

• Increased danger of digital authoritarianism.

• Work in diplomacy, such as efficiency of consular work.

• Normative challenges to existing values and considerations.

• The race to be leading AI development, research and implementation, including through data collection.

• The perceived need for international governance of AI technologies.

Narrowing Down What and When Constitutes “Impact”

Participants repeatedly debated the timeframe that applies to the question of AI’s impacts on geopolitics. The contemporary landscape differs from potential future scenarios given the rapid development of AI technologies. The assessment of what kind of development to expect and the capabilities AI technologies may develop, however, strongly depends on the taken approach: tech-optimist or doomsday scenario. The timeline envisaged is also relevant. One participant pointed out that “the actual impacts in the short, mid- and long term are somewhat unclear”. Even the question of what timeframe constitutes “the future” or “long term” is challenging.

Timeline considerations also tied to questions on whether to focus on long-term, potentially existential, risks that largely rely on speculation or to focus on AI’s contemporary impact and potential harm. One academic pointed out that this discussion was now highly politicised, tied to narratives of US–China competition and other broader trends. They argued, however, that long- and short-term impact cannot be disintegrated so easily but that these questions are instead closely tied to political considerations underlying them.

The political implications that follow from the narrative of distinguishing between long- and short-term impact were picked up by other participants. One scholar argued that such distinction is not contradictory. Nonetheless, it is vital to bridge the gap between various communities, each focusing on different types of risks (for example, the national security community and the military community).

Especially when discussing long-term and existential risk of AI technologies, academics stressed the need to consider where the information underlying assumptions comes from. Who circulates what information with what kind of intentions? What vision of technology are they trying to sell? One participant found that this increased politicisation is making their job harder. Another academic also pointed to the lack of diversity among many technology companies. Many of these companies hold a lot of information and knowledge. They shape the discourse on the impact of AI, especially when it comes to discussing potential long-term and existential risks. They may do so in a one-sided way. Another participant added that technology companies often even determine what is labelled as AI, thereby holding additional power.

All of these points on the different perspectives, the timeframe applied and the information underlying any assumptions shape the understanding of AI’s impact on geopolitics.

The Approach of International Relations Scholarship to AI

The discussion identified several areas of the impact of AI on geopolitics that are – to varying degrees – addressed in International Relations scholarship. For example, one participant’s introductory remarks argued that International Relations scholars primarily address the impact of AI by focusing on four areas of study. First, the influence of AI on the balance of power is primarily addressed in areas such as war studies or strategic studies. Second, issues of AI governance are analysed by a more diverse set of scholars driven by a shared sense that some form of global AI governance is needed. Third, the same scholar further remarked that ethics questions are often primarily looked at within a military context but should be considered more broadly. Finally, disinformation and social media are widely discussed in International Relations scholarship, particularly considering the impact of deepfakes, but research in this area is often inconclusive.

The discussion on AI’s impact on geopolitics underlined the implications that different framings have, whether on the temporal framework applied, the community and the political environment that discusses AI’s impact on international relations, or the source of information such assessment is based on in the first place. These challenges, particularly the question of timeframes, were raised throughout all stages of the workshop.

Part Two: Learning from Other Technologies

AI technologies are by no means the first technological invention to impact international relations. From nuclear to space, and military inventions to maritime technologies, technology impacts many areas of interstate relations. Workshop participants therefore also debated how far AI technologies and their impact on geopolitics resemble that of other technologies or whether they pose unique, unprecedented challenges.

How Different is AI from Previous Technological Inventions?

Assessments of how far AI is “a gamechanger” differed among the participants. Again, varying timeframes applied to the question. On the one hand, some considered that AI, at least in the current context, primarily amplified existing challenges or power dynamics without fundamentally challenging interstate relations. Some participants tied the impact of AI on international relations to other technologies (for example, to quantum sensing or information communication infrastructure [ICT] and cyber technologies), pointing to possible similarities and interdependencies. One participant also argued that “AI has far more in common with earlier technological inventions than is often implied in the existing literature”, at least in its current form. This could, in theory, change if artificial general intelligence comes into existence.

On the other hand, some participants found that the impact of AI is already groundbreaking. They stressed its ability to replace human factors as – unlike other technologies – it is a decision-making technology with “a degree of agency”. AI’s ability to not just replace motor skills – as was the case with previous technologies – but also “chiefly cognitive skills” was perceived as a decisive difference. As such, AI has the ability to change the quality of decision-making, with implications for the nature of international society and relations.

More abstractly, as International Relations as a discipline concerns the study of interstate relations, not humans, AI fundamentally challenges this assumption, according to one discussant. Here lies the challenge to International Relations as a discipline. At the same time, it can also add to the discourse: while computer scientists focus on individual intelligence, International Relations scholars focus on collective social intelligence because the discipline is the study of groups, not the individual. As such, the field’s scholarship can make a meaningful contribution to often technology-dominated discussions on how the impact of AI on international relations is understood.

What Lessons Were Learned from Other Technologies and Their Impact on International Relations?

On lessons learned from other disciplines, including on how to deal with increased hype around new technological inventions, one participant remarked that in the area of space technology, camps form around “futurists” and those paying more attention to contemporary developments. While futurists primarily focus on potential technological inventions in the more distant future, possibly developed by engineers without guiding policy and strategy, those who are perhaps more grounded in pragmatism focus on the technologies that are actually being deployed. The participant remarked that a lesson learned from the space context is to “look away from the shiny stuff” and instead focus on where mass investments are being made and mass adoption is being advanced. A similar shift in attention in an AI context would contribute to moving away from the contemporary hype around AI and to guide non-technicians in navigating the technological landscape and seeing the bigger picture.

Interesting lessons were also drawn in comparison to the development of nuclear technologies. A participant noted that the ownership of the development of AI – unlike that of nuclear weapons – lies with technology companies rather than states. This, for them, challenges the state-centric assumptions often underlying International Relations theory.

Another comparison was made to the nuclear domain. Like the sanctions there, one participant expected future geopolitical considerations on AI development would likely include further sanctioning of semiconductors. This is already a powerful tool of US foreign and technology policy.

Others stressed that AI technologies cannot be seen in isolation and that they do “not exist in a social vacuum”. Instead, AI technologies must be seen as a collection of technologies which are also linked to broader questions such as supply chain issues or labour market dynamics.

While participants’ observations of how far AI technologies constitute a game changer or resemble previous technologies’ impact on international relations differed, some interesting parallels and differences were drawn. These highlighted the value of situating our understanding of AI in the broader context, both in relation to parallel political developments and dynamics around other technologies.

Part Three: Applying Existing Theories and Frameworks to AI and the Need for a New Concept

At the start of the workshop, participants were asked to place themselves on a spectrum ranging from “no new framework needed” to “new framework needed” to analyse AI’s impact on geopolitics (see Figure 1 for a rough indication). The allocations are merely indicators replicated from a whiteboard in the room and do not follow a scientific measurement. Nevertheless, they provide an overview of the different sentiments represented during the workshop.

▲ Figure 1: Do We Need a New Framework to Understand AI’s Impact on Geopolitics?

The Need for a New Framework to Capture the Impact of AI on International Relations

As demonstrated in Figure 1, participants offered a wide range of perspectives and arguments on whether a new analytical framework in International Relations scholarship is needed to adequately capture and explain the impact of AI on international relations.

Some thought that a new framework is needed to go beyond the descriptions of aspects of AI’s impact offered by existing theories and concepts. A new framework would especially need to address AI’s ability to perform cognitive tasks and its impact on political decision-making. One participant also deemed a new framework necessary to move away from the dominant perception of AI as an arms race between big powers, overlooking smaller and medium states as well as private technology companies, which are at the heart of AI development. Whether such a new framework would rely on existing concepts that needed updating and enhancing, including from an interdisciplinary perspective, or whether a new, comprehensive theory should be developed remained subject to debate.

Others pointed to the many theories of International Relations scholarship that already apply to technological inventions more broadly. For example, the framing of the governance and technology lag, describing how technology develops faster than its respective governance, applies to an AI context. Another participant pointed to critical norm theory as a way to build on existing theoretical thinking on how AI technologies shape international norms. For specific areas such as trade and international political economy, the new trade theory was named as an example of an existing theory, adding insights to AI’s impact on geopolitics. Similarly, the theory of organisational reputation was given as an example of how an existing theory can “provide insights into why agencies react to and regulate disruptive innovations”. One scholar added that existing concepts from other disciplines such as neuroscience, psychology and philosophy are also useful tools in contributing to explaining AI’s impact on international relations.

One Theory to Rule Them All?

The discussion arguing in favour of one new concept or against it was far from binary. Instead, it was often a matter of the degree to which existing theories can adequately capture AI’s impact. Participants further questioned whether there can be one theory to address AI’s impact on international relations in the first place. While one participant pointed out that the big theories – realism, constructivism and institutionalism – each offer “valuable insights into AI and the impact it may have on geopolitics”, others stressed the move away from “the big ‘isms’”. Instead, there are many more nuanced, mid-level theories aiming to address aspects of International Relations.

Indeed, one scholar argued that “no single theory, concept or framework can comprehensively and adequately describe, analyse and reflect on the various impacts of AI. Each theory in International Relations can, however, illuminate certain aspects, risks and opportunities of AI, but none can on its own discuss, analyse and reflect on it in a comprehensive manner”. Others agreed, stating that they did not think “that one single theory, concept or framework can fully capture the impact of AI technologies as they all have their own comparative strengths and weaknesses. For this reason, a multiple theory, concept or framework is the most analytically productive”.

Some noted that a comprehensive approach would be desirable to accomplish a new, overarching theory, acknowledging the tremendous challenge of developing such a theory. Others argued in favour of basing work on existing theories and updating them in line with new developments.

What Tool for What Purpose?

The discussions on the need of a new framework or theory repeatedly linked back to questions of purpose, scope and intention. Participants discussed both what aspect of AI technologies a concrete framework should focus on but also what the purpose of International Relations scholarship is more fundamentally. Again, defining purpose also tied to the question of timeframes – whether a new theory would address the impact of AI on international relations now or in the future. One participant argued that, as it stands, existing concepts “have a lot to say” but that it is harder to predict future developments and how existing concepts would be able to keep up.

With respect to the purpose of a new framework, participants asked why a framework is needed, and whether it is to better understand AI’s impact on geopolitics as opposed to drawing attention to previously unnoticed phenomena. One participant argued that, in some sense, International Relations scholarship “had a technology problem” – while technology was part of many considerations, it was never conceptualised as such. This status quo is now challenged, as new developments in AI technology arguably have the potential to fundamentally challenge how International Relations theories work.

This linked to the broader and fundamental question of the purpose of International Relations theory. Participants critically discussed whether it is the purpose of International Relations theory to predict the future; a challenge some considered too great given that the “past and present are already difficult enough” to conceptualise. Others felt that it is indeed the idea of the discipline to predict geopolitics but that it has also been historically poor at doing so. This also raised the question of how far International Relations theory is contributing to preventing war or harm and whether that is its purpose and, if so, whether it needs to get better at doing so.

Thus, how International Relations scholars understand their discipline and interpret existing theories heavily influenced their assessment of whether a new framework is necessary or feasible and, if so, what purpose it might serve.

Conclusion

The workshop identified that whether participants considered a new framework necessary to adequately capture AI’s impact on international relations was influenced by three key factors.

1. What experts considered AI’s impact to be in the first place – and how unprecedented they found it compared to previous technological inventions.

2. How far they believed existing theories in various fields can capture such impact and whether a single theory could ever do so.

3. Experts’ assessment of how useful existing theories are also depended on what they considered the objective and purpose of both a new theory and International Relations scholarship more widely to be.

The conversation repeatedly returned to themes such as the applicable timeframe to define the impact of AI, the definition of AI, and the bias of available information and their implications on scholars’ understanding of AI. These factors, the assessment of which varied depending on participants’ areas of specialism, were often considered underlying questions that need to be answered before a suitable framework can be developed.

Further research could provide a fuller and more systematic understanding of AI’s impact on geopolitics but also help identify concrete examples of how far existing theories already capture various aspects of the impact of AI on geopolitics. Research could also provide a more detailed analysis of parallels and differences between AI and previous technologies and their impact on international relations.

Annex: Survey Questions

1. What impact do AI technologies have on international relations?

2. To what extent does this resemble or differ from previous technological inventions?

3. Can existing theory, concepts or frameworks in your field adequately describe the impact of AI technologies? (Please expand and include field and theory.)

4. If applying existing theories does not work, why not? And what do new analytical frameworks need to offer to successfully reflect the impact of AI technologies on geopolitics?

Pia Hüsch is a RUSI Research Fellow in cyber, technology and national security. Her research focusses on the impact, societal risks and lawfulness of cyber operations and the geopolitical and national security implications of disruptive technologies, such as AI. Pia’s other research interests include the governance of cyberspace, election interference, cyberwarfare, and the relationship between law and technology, including cyber and AI.