Strategy and Grand Strategy in an Age of Artificial Intelligence

Summary
The Institute for Global Politics (IGP) at Columbia University’s School of International and Public Affairs (SIPA) convened an intimate group of leading experts for a closed-door convening to tackle cutting-edge academic and policy questions regarding the implications of Artificial Intelligence (AI) for strategy and grand strategy. The purpose was to “bridge the gap” between scholars working in this area and practitioners actively grappling with these issues.

This convening focused on two core questions relevant for both academics and policymakers:

• What do innovations in AI suggest about the future of US military strategy? What is the best way to fight and win wars in the age of AI?
• What do such innovations suggest about the future of US grand strategy? What is the best way for the US to secure itself in the age of AI?

The relationship between strategy, grand strategy, and technological innovation is at the core of much of the literature on international security and is an enduring priority for policymakers. The purpose of the convening was to begin to outline how recent advanced of AI on the battlefield should inform future military strategy, as well as how American grand strategy should incorporate AI technologies more broadly. Participants identified potential policies the US should put in place to achieve the desired goals of strategy and grand strategy, and explored potential risks and confounders.

The convening tackled these questions through two lenses. First, the group explored how AI is changing military strategy through the lens of the 2022 Russo-Ukrainian conflict, where AI-enabled military capabilities have played a significant role on the battlefield. Then, the group assessed how American grand strategy should grapple with technologically driven changes in the balance of power, with a focus on the US-China rivalry.

The objective of the roundtable was to develop an understanding of the current state of the field, identify pathways for future research and policy work, and help foster a community of academics and policymakers working in this area.

The convening took place under Chatham House rules.

Discussion: Part I
Part I of the roundtable centered around the lessons of AI on the battlefield from the Russo-Ukrainian conflict. The discussion kicked off with a conversation about how highly anticipated and hyped innovations in AI have been integrated into battlefield technologies, particularly in the context of the Ukraine conflict. AI has enabled the Ukrainian army to ingest and analyze large amounts of data from diverse sources—such as satellite imagery, UAVs/drones, social media content, radio signals, and crowdsourced-geotagged photos—to collect intelligence about Russian troops' locations and movements and improve precision targeting. At the same time, the Russian military has also deployed AI-enabled capabilities on the battlefield, including uncrewed and autonomous systems.

The participants discussed the tactical, operational, and strategic implications of these military applications of AI technologies. In the Ukraine conflict, AI has played a critical role in pattern recognition; intelligence, surveillance, and reconnaissance (ISR); targeting; maneuver; and military logistics. Innovations in computer vision and automated targeting have been significant, with AI-guided systems demonstrating an ability to identify and track targets autonomously, maneuver to evade countermeasures, and successfully hone in on and hit desired targets. Data fusion and predictive analysis have also been essential in enabling intelligence platforms to integrate information from satellites, signals intelligence, open-source information, telecommunications, social media, and so on to enable faster decision-making. However, they also noted that implementing AI on the battlefield was in some cases constrained by the availability and robustness of infrastructure to enable military organizations to aggregate data, securely transfer it, and then update it to produce high-fidelity battlefield data at the tactical edge.

One core theme that emerged from the discussion: there is a great deal of hype around military AI applications, some of which are based on untested assumptions or gaps between the expectations and enthusiasm of innovators in the AI technology sector and the knowledge and experiences of operational commanders. AI has demonstrated considerable tactical value on the battlefield in Ukraine. Specifically, the experts identified two broad-ranging tactical benefits of military AI. The first is improving situational awareness for frontline commanders through integrating data from diverse sources. This enables them to rapidly adapt to changing battlefield conditions, anticipate and get ahead of adversary maneuvers, and seize the initiative. The frontline in Ukraine has been characterized by proliferating sensors in a way that is distinct from any previous conflict. These sensors enable the collection of a vast range of data that can inform battlefield operations, including not only imagery (such as from overhead reconnaissance satellites) and acoustic (such as detecting aerial threats through analyzing sound data), but also text data from social media platforms and text message communications.

The second tactical utility of AI is enabling autonomous systems, especially last-mile solutions—which allow for an uncrewed platform, such as an uncrewed aerial vehicle, to engage a target autonomously, without human control. Uncrewed aerial vehicles and other drones have been a defining feature of the Ukraine conflict, and there is a perception that these capabilities are relatively easy to use. However, the reality is that there is a good deal of skill involved in using drones effectively. AI tools have been critical in reducing the level of skill necessary to operate first-person view drones and evade adversary countermeasures like electronic warfare (EW).

Despite these clear tactical applications of AI tools, participants agreed that it is far more challenging to identify the implications of AI across echelons and at scale. While the tactical benefits are more tangible, it is far less clear how these tactical implications aggregate to the operational and strategic levels. For one, much of the debate around the battlefield implications of AI does not distinguish between the benefits and limitations of AI across different warfighting functions, capabilities, and operational domains. Additionally, scaling AI in contested environments has been extraordinarily challenging. EW capabilities, degraded communications, and the swift pace of adversary countermeasure development has posed serious challenges for capitalizing on AI at scale. This has made innovation-driven advantages on the battlefield fleeting. Another limitation has been the prospects of UAV swarming. While there has been considerable bluster and optimism about the feasibility and impact of drone swarms, there is not clear, compelling evidence from the Ukraine conflict of fully autonomous drone swarms on the battlefield. Moreover, when it comes to operational and strategic level issues, such as command and control and decision-support tools, AI tools have been less impactful. The reality is that AI decision-support tools are still too brittle for strategic decision-making in complex, high-stakes combat scenarios.

Another recurring core theme was military innovation. The participants noted that the development and integration of AI tools into military operations in Ukraine occurred organically and relatively quickly, through bottom-up mechanisms, rather than through a more deliberate, top-down process or comprehensive strategy. This is not surprising, given the different nature of wartime versus peacetime military innovation in general and the existential stakes Ukraine perceives in this war in particular. In turn, Ukraine’s defense sector has benefitted from opportunities for entrepreneurial involvement and an embrace of open systems and natural prototyping. The private sector, including both US/Western technology companies as well as Ukrainian startups, have played an enormous role in providing Ukraine with critical technologies, capabilities, and services. But this poses both risks and opportunities, and raises policy challenges. In terms of opportunities, private-sector led innovation has enabled faster fielding of battlefield systems via dual-use technologies. The private sector also cultivated a grassroots innovation ecosystem in Ukraine, where private entities develop unique solutions tailored to specific, often narrowly defined, battlefield challenges or applications. A particular feature of this kind of innovation has been an embrace of open systems that enable frontline commanders and soldiers to experiment with and developed tailored tools that are matched to their specific requirements. But the role of the private sector is also creating potential risks. One is that it is blurring the lines between civilian and military actors, which raises questions about what constitutes a legitimate target and may be creating potential escalation risks. Another is that private companies engage and provide support at their discretion but, especially for non-Ukrainian entities, this can change due to new considerations, interests, risks, and other factors, raising questions about the reliability of private sector support over the long term. The involvement of private actors, especially non-Ukrainian commercial entities, also creates challenges in terms of accountability and the chain of command.

The participants then turned to the implications of Ukraine’s bottom-up approach to battlefield innovation and adoption of AI tools contrasts with the current US defense acquisitions approach. In the United States, defense acquisition is a purposefully slow and deliberate process. This is in large part due to an implicit assumption among US policymakers that the US will retain a technological advantage over its adversaries and, therefore, can afford to take a more deliberate approach. However, this assumption may no longer hold in the current international environment, as the US faces great power rivals who are investing heavily in the development and adoption of military AI innovations, and as the pace of innovation overtakes the defense acquisitions process.

One participant pointed to a puzzle in US adoption of military AI. On the one hand, the US military has been investing considerable resources to drive innovations in the development of AI-enabled military capabilities. However, it struggled with systematic adoption at scale. This points to a critical distinction between innovation—developing the capability—and adoption—integrating it into organizations, doctrine, and processes, either through adjusting exiting ones, developing new ones, or both. Programs to develop and adopt AI solutions are siloed across various units and service branches, but there is no comprehensive effort to adjust the defense ecosystem to enable to coherent and systematic integration of AI tools across the defense enterprise at scale. Clearly, Ukraine and the US face vastly different challenges, resources, constraints, and interests. But Ukraine’s ability to rapidly integrate and accept some measure of risk in the adoption of advanced AI capabilities is a useful vehicle for shedding light on the limitations of the US defense acquisitions process. While there have been pockets of decentralized innovation and experimentation across different parts of the US military, this kind of bottom-up innovation has not resulted in permanent institutionalization or demonstrated capabilities that can be scaled. AI is not currently integrated across the joint force. Most participants agreed that the process of adopting AI technologies into the US military should be deliberate, but they were divided on whether decentralized experimentation is a productive method for achieving adoption at scale.

Discussion: Part II
Part II of the roundtable focused on the implications of AI for US grand strategy. The general-purpose nature of AI makes it well-suited to address questions at the level of grand strategy, as AI speaks to the full complement of national power. Both the United States and China recognize the strategic potential of AI to shape the future balance of power, with implications not only for generating military capabilities, but also for economic growth, alliances, intelligence collection, and other facets of power. Future American grand strategy will need to address a range of difficult questions involving AI, from the fundamental question of defining objectives, to articulating clear priorities that can drive investments in resources across the instruments of national power, identifying areas of comparative advantage and disadvantage, and addressing risk management, among other issues. Several core insights emerged from this part of the discussion.

One important finding was that AI is becoming a core driver of national power and this should have consequences for US grand strategy. However, in the post-Cold War environment the US has struggled with developing a coherent grand strategy, with a lack of consensus around competing US grand strategy paradigms, such as restraint (or neo-isolationism), primacy, and liberal internationalism. There has been little strategic thinking to articulate how AI fits into existing paradigms, or whether an entirely new grand strategy paradigm is needed to account for the role of AI in the contemporary age. In particular, beyond narrow questions about the technical skill sets and capacities needed to pioneer the development of cutting-edge AI capabilities, a cohesive grand strategy for the AI age should take into account the supporting sectors and infrastructure that will be imperative to sustain continued AI growth. This includes access to critical resources (such as rare earth minerals), power, and computational infrastructure. These themes are prevalent in both the current and previous presidential administrations’ strategies, but have not yet been translated into a cohesive worldview.

A related theme was examining the nature of great power competition between the US and China through an AI lens. Participants brought up that China’s grand strategy for AI is on far firmer footing than the United States’. Beijing’s approach to becoming the world’s leading AI power by 2030 is focused on a holistic, state-driven strategy to drive investment and innovation in AI, build a domestic cadre of AI talent through education, build necessary infrastructure, collect data at scale, and so on. Innovation-driven development is a priority for the Chinese government, and military-civil fusion is the mechanism for this development, which connects military and economic power. Beijing’s centralized approach to grand strategy for AI involves investing heavily in infrastructure and supporting sectors that are essential for AI development, like deep sea mining and the massive expansion of data centers. With a very different political system, the US will by necessity take a more collaborative and decentralized approach to public-private collaboration and the cultivation of AI power. The participants also discussed DeepSeek, a Chinese AI startup that recently disrupted the tech sector when it introduced an AI assistant that was developed at a far lower cost than US competitors such as ChatGPT. DeepSeek illustrates how, if the US restricts the diffusion of AI technologies, China will likely find a way to produce its own alternatives and make it available to others. It also demonstrates that frontier technology will not necessarily be decisive in determining whether the US or China “wins” the AI race, as “good enough” AI tools may be sufficient for most applications.

The discussion turned to potential policy prescriptions for the United States. Linking the first and second parts of the conversation, the experts recommended that the US focus on developing solutions and improving processes for the adoption of AI at scale, rather than singularly focusing on driving innovation, as a means of gaining strategic advantage relative to China. Another core recommendation was to adopt an AI-centered grand strategy that prioritizes developing and gaining access to the infrastructure and resources needed to support AI innovation and adoption. The US should drive investments in critical infrastructure for AI (i.e., data centers) and supporting sectors (i.e., semiconductor manufacturers, cloud service providers, utilities, hardware manufacturers, etc.). With respect to enabling innovation the private sector, the US should more robustly fund foundational research in AI and cultivate innovation ecosystems, similar to how the US government partnered with leading research institutions to help create the internet. The government should also promote greater regulatory clarity and data access frameworks to empower private innovation. Another critical policy issue entails improving the domestic talent base through investing in STEM-based education pipelines and promoting programs to recruit and retain top AI talent from around the world.

Additionally, participants advocated that the US should more robustly participate and assert a leadership presence in multilateral and multi-stakeholder discussions, especially those that are focused on norms, standards-setting, and global governance around AI. These include international organizations that serve as platforms for coordination, such as the OECD, G7, and UN, and technical standards-setting and regulatory bodies such as the International Telecommunication Union and the International Energy Agency. A related critical area the US should pursue is the development of confidence-building measures—even with adversaries and competitors—for those areas where there are overlapping interests around restraints and transparency about AI. For example, both the US and China have a shared interest in avoiding inadvertent escalation stemming from the use of military AI capabilities (such as in the area of nuclear command and control). Another type of confidence-building measure could be for the US and China to pursue an autonomous incidents agreement, similar to the 1972 Incidents at Sea Agreement between the US and the Soviet Union during the Cold War that sought to avoid inadvertent escalation and facilitate crisis stability in the context of naval interactions. Such confidence-building efforts in the AI age could present a viable avenue for cooperation, despite conflicting interests in other aspects of AI governance. Finally, the participants debated the appropriate economic levers the US should employ to solidify US advantages in AI innovation. Some advocated for export controls—regulations that restrict the exporting of capabilities, services, or data that could have deleterious national security consequences—and other economic measures. However, participants also cautioned that the US needs to have a more precise concept of export controls and to use them as a means of buying time for scaling AI adoption and allowing US companies to innovate to take the lead in the AI race against China. Broad economic measures could produce unintended and counterproductive effects, such as driving further domestic innovation within China.

Overall, the participants largely agreed that AI will have significant implications for the future of warfighting and grand strategy more broadly, but also cautioned that some of the current hype about the potentially transformative or even revolutionary implications of AI may be overblown. Further research is needed to carefully evaluate the effects of AI at the strategic level, as well as to develop credible measures of AI power to assess how AI may be changing the distribution of power in the international system. Moreover, the experts agreed that AI is giving rise to a broad and diverse range of policy challenges across all facets of national power. In turn, it will be imperative for the US to take a deliberate and carefully considered approach to develop a clear grand strategy that accounts for this general-purpose technology.