The AI Wave in Defence Innovation

Michael Raska and Richard A. Bitzinger assembled this volume at the S. Rajaratnam School of International Studies in Singapore in the years that followed the United States Defense Department’s Joint Artificial Intelligence Center, the announcement of China’s “intelligentization” doctrine, and the first wave of national AI strategies issued in capitals from Paris to Tokyo. By the time Routledge published the book in 2023, the comparative landscape had thickened to the point where single-country treatments no longer captured what was happening. Raska, a senior fellow whose own work has tracked Asian military innovation closely, and Bitzinger, who has spent decades on the defence-industrial side of strategic studies, gathered eleven contributors with country and regional expertise. The result joined Routledge’s Modern Defence and Strategy series, sitting alongside earlier volumes on cyber, missile defence, and great-power competition.

A conversation the book joins is one that English-language defence literature had been holding largely through an American lens. Books such as Paul Scharre’s “Army of None,” Kenneth Payne’s “I, Warbot,” and the Brookings and CNAS reports of the late 2010s had framed military AI through the prism of US-China rivalry and Pentagon programmatics. What Raska and Bitzinger contend, in their introduction and through the chapter selection, is that the diffusion of AI into armed forces is not a single race emanating from Washington. National strategies are inflected by industrial bases, civil-military relationships, alliance commitments, threat perceptions, and procurement cultures that vary widely. A common stack of technologies — neural networks, autonomy software, sensor fusion, data infrastructure — lands very differently in different political economies. The speed at which a force can move algorithms into command, intelligence, surveillance, reconnaissance, and weapons depends as much on legal posture and export-control regimes as on the underlying code.

Built around that thesis, the volume opens with three chapters on the great powers. A United States chapter walks through the Third Offset Strategy under Robert Work, Project Maven and the disquiet at Google that followed, the establishment of the Joint Artificial Intelligence Center and its successor the Chief Digital and Artificial Intelligence Office, and the slow grind of bringing predictive maintenance, computer vision for ISR, and decision-support tools through a procurement system designed for hardware platforms. A China chapter reconstructs the People’s Liberation Army’s doctrine of “intelligentized warfare,” the military-civil fusion arrangements that route work between firms such as SenseTime, iFlytek, and Megvii and the PLA’s research institutes, and the Strategic Support Force restructure of 2015 that placed information warfare alongside the traditional services. A Russia chapter is the most cautionary of the three: it documents ambitious doctrine — Vladimir Putin’s much-quoted 2017 remark that the leader in AI would rule the world — against an industrial base hollowed out by sanctions, brain drain, and dependency on imported semiconductors, with the war in Ukraine after 2022 throwing the gap between rhetoric and capability into hard relief.

From there the volume turns to what is in many ways its most useful section: a chapter each on Israel, India, South Korea, Australia, France, Germany, Singapore, and Japan. An Israeli chapter draws out the deep loop between the Israel Defense Forces’ Unit 8200, the country’s commercial AI sector, and prime contractors such as Israel Aerospace Industries, Elbit Systems, and Rafael Advanced Defense Systems. It covers Iron Dome interceptor cueing, the loitering munition family that runs from the Harop down to smaller man-portable systems, and the targeting-recommendation tooling that has gone by names such as Habsora and Lavender in subsequent press accounts. A picture emerges of a small state that treats its civilian software workforce as a strategic reserve and accepts a degree of operational risk that larger democracies would not.

An India chapter is a study in the gap between ambition and execution. It walks through the Defence Research and Development Organisation’s AI roadmap, the Atmanirbhar Bharat indigenisation drive, the establishment of the Defence AI Council and the Defence AI Project Agency, and the procurement reforms intended to pull start-ups into the supplier base. The chapter is candid about the bottlenecks: a thin domestic semiconductor base, dependency on Russian and now increasingly Western platforms whose data does not flow easily to Indian developers, and a civil-military culture in which the services have only recently begun building software capacity in-house.

South Korea is presented as one of the more striking adopters. A chapter traces the Defense Reform 2.0 programme, the Defense Acquisition Program Administration’s push toward indigenous autonomous systems, the role of Hanwha Aerospace, LIG Nex1, and Korea Aerospace Industries in unmanned ground and air platforms, and the demographic logic — a shrinking conscript pool — that pushes Seoul toward automation more aggressively than its allies. The K-defense export surge that took off with the Polish K2 tank and K9 howitzer deals is read here as the commercial flywheel that helps fund AI integration into the next generation of Korean systems.

Australia’s chapter centres on AUKUS, the 2023 Defence Strategic Review, and the Loyal Wingman programme that became the MQ-28A Ghost Bat — Boeing Australia’s collaborative combat aircraft and the first military aircraft designed and produced on Australian soil in more than half a century. The chapter handles the tension between a defence budget large by middle-power standards but small by great-power standards, and an alliance structure that gives Canberra privileged access to American and British technology in exchange for a tighter strategic posture in the Indo-Pacific.

A France chapter treats the Agence de l’Innovation de Défense, established in 2018, as the institutional centrepiece. It traces the AI strategy document issued by the Ministry of the Armed Forces, the Loi de programmation militaire’s spending allocations, and the careful French insistence on a doctrine of “responsible AI” that retains a human in the loop on lethal decisions. Dassault, Thales, and Safran appear as the national champions, with the start-up firm Preligens — previously known as Earthcube — singled out as the country’s most prominent specialist in geospatial-intelligence AI. The chapter notes France’s sustained effort to keep European AI defence work from being fully absorbed into either an American or a transatlantic-NATO orbit.

Germany is treated with that country’s own ambivalence. The chapter spans the period before and after the February 2022 Zeitenwende speech, in which Chancellor Olaf Scholz announced the hundred-billion-euro special fund for the Bundeswehr. It documents the Cyber and Information Domain Service, the Bundeswehr’s tentative experiments with predictive maintenance and battlefield-awareness tools, and the rise of Helsing, the Munich start-up that has become Europe’s most visible defence-AI company. The chapter does not minimise the cultural and legal headwinds — the Bundestag’s parliamentary control of armed force, the post-1945 reticence about lethal autonomy — and treats them as parameters the German defence-AI ecosystem has had to design around rather than obstacles that can be dissolved.

Singapore receives its own chapter and benefits from the editors’ proximity. An argument runs that the Singapore Armed Forces have built one of the more coherent small-state digital-force concepts in the world: a Digital and Intelligence Service stood up as a fourth service in October 2022, a Defence Science and Technology Agency that has long operated as the country’s technical procurement arm, and a tight integration with the Republic’s universities and with firms such as ST Engineering. The chapter argues that Singapore treats its compactness as a feature, allowing the country to iterate on integrated data architectures that larger forces struggle to impose across legacy stovepipes.

Japan closes the country tour. A chapter covers the December 2022 National Security Strategy and the accompanying Defense Buildup Program, the decision to acquire long-range strike capabilities including Tomahawks, the new Joint Operations Command structure, and the role of Mitsubishi Heavy Industries, NEC, and Fujitsu in defence-software contracts. The Global Combat Air Programme with Britain and Italy, announced in late 2022, appears as the most consequential AI-relevant collaboration on Tokyo’s docket. The chapter handles Article 9 and Japan’s evolving export-control reforms with care, treating them as the constraints that have shaped what Japanese defence AI is allowed to be.

Closing chapters return to the strategic level. One looks at alliance dynamics, focusing on how NATO, AUKUS, the Quad, and the JAIC’s international engagements are creating multi-tier interoperability problems that machine-readable doctrine and shared data standards have only begun to address. Another treats technology-transfer regimes — the Wassenaar Arrangement, the United States’ export-control reforms, the European Union’s Dual-Use Regulation, and Japan’s evolving control list — as the increasingly contested gates through which AI-relevant components and software flow. A final analytical chapter takes up the recurring tension between moving fast on AI capability and retaining the testing, evaluation, verification, and validation standards that allied operations are expected to meet. Contributors do not paper over the gap: they treat it as the defining governance problem of the next decade of military AI.

Reception within the field has been measured. The volume has been picked up in graduate seminars on military innovation and in policy shops in Europe and Asia that needed a country-by-country reference rather than another single-author argument. Reviewers have generally welcomed the comparative breadth and flagged what is inevitably missing — there is no dedicated chapter on the United Kingdom, on Taiwan, on Turkey, or on the Gulf states, all of which have become more relevant since the book’s manuscript closed. Specialists on autonomous weapons have noted that the volume does not engage deeply with the Convention on Certain Conventional Weapons debate or with the philosophical literature on meaningful human control; that work lives in Heather Roff, Paul Scharre, and the Stop Killer Robots coalition’s output, and the editors plainly chose to leave it to those authors. The book’s empirical chapters have aged variably: the United States, French, and German chapters in particular were overtaken by events as the Ukraine war reshaped European defence spending and the Pentagon stood up the Chief Digital and Artificial Intelligence Office.

For a reader trying to navigate the field today, the volume sits as a comparative-policy reference. It pairs naturally with the doctrinal literature on autonomous weapons — Scharre’s “Army of None” and “Four Battlegrounds,” Christian Brose’s “The Kill Chain” — and with the more technical defence-AI engineering work coming out of CSET, RAND, and the academic computing community. Where those books either argue a position or describe a single national programme in depth, Raska and Bitzinger’s collection answers a different question: what the international system actually looks like when one looks across the middle powers and the regional players who are doing the work without the bandwidth to broadcast it. Anyone trying to understand why a Korean, Singaporean, Australian, or French procurement decision lands the way it does — which firms get funded, which capabilities get prioritised, which exports get blocked — will find the country chapters the obvious place to start.

What is likely to age well in this book is the framing. A proposition that military AI is a diffusion problem rather than a race, and that the diffusion is shaped by industrial and political structures specific to each adopter, is the kind of analytical move that holds up regardless of which particular capability is the headline of the year. What will date faster is the empirical surface — the specific programmes, budget figures, and personalities — which in a fast-moving field will need refreshing well before the decade is out. Raska and Bitzinger appear to have written the book knowing this, structuring it so that the comparative method survives even as the case material is overtaken. That is the discipline of an academic reference work, and on its own terms it is the volume’s most useful contribution to a field still working out what it is going to look at next.