Contribute

Responsible Procurement of Military Artificial Intelligence in Australia

A Window of Opportunity

Authors: Netta Goussac and Vincent Boulanin

Introduction

States are rushing to adopt artificial intelligence (AI) in the military domain. Australia’s 2026 National Defence Strategy, for example, states that ‘Defence will demonstrate, develop, certify and integrate artificial intelligence into capability safely and rapidly’.[1] Converging factors are driving this urgency. Strategic competition between the major powers has intensified, with the United States of America, China and Russia each investing heavily in military AI as a source of competitive advantage.[2] Recent conflicts have reinforced the perception—first expressed by the US in its 2016 ‘third offset strategy’[3]—that AI will be central to the conduct of modern warfare. In the 2020 Nagorno-Karabakh conflict, AI-enabled loitering munitions played a significant role in Azerbaijan’s victory; since 2022, Ukraine’s response to Russia’s invasion of its territory relied on innovative uses of robotics and AI; Israel has relied on AI extensively to plan and conduct military operations in its conflict with Hamas in Gaza since 2023; AI was used by Israel and the US to identify targets and compress decision timelines in the 2026 conflict with Iran.[4]

How states should pursue the adoption of AI capabilities in the military domain remains, however, a challenging question, as they face two difficult and interrelated issues when it comes to the procurement of such capabilities.

The first concerns the need for responsibility. States need to ensure the AI capabilities they procure are safe, secure, trustworthy, lawful and aligned with the political commitments they have made around responsible military AI at the national or international level.

The second pertains to the need for speed. In a deteriorating geopolitical environment many states are under pressure to accelerate their adoption of AI in the military.[5] AI is perceived as offering an indispensable strategic and operational advantage. This narrative is echoed and amplified by industry actors, who position AI capabilities as essential to extracting actionable insights from ever-growing flows of intelligence, surveillance and reconnaissance data, and deploying more autonomous weapon systems.[6] A chorus of voices from militaries and industry alike have called for reform of military procurement processes to allow more rapid acquisition and fielding of AI capabilities. States are still grappling with finding a balance between their desire to streamline procurement pathways to facilitate the adoption of AI and the commitment to preserve the elements that ensure such capabilities will be procured—and ultimately used—responsibly.

This paper explores how Australia could reconcile the perceived need for speed with the need for responsibility in the procurement of military AI capabilities.

Australia’s planned reform of procurement is in part driven by the desire to accelerate uptake of new technologies. But it also provides a window of opportunity for Australia to implement its high-level policy commitments and international legal obligations around responsible military AI.

The paper is primarily based on desk research and interviews that the authors conducted on the responsible procurement of military AI in the context of a project funded by the Department of National Defence of Canada and the Ministry of Defense of the Kingdom of the Netherlands. The paper is also informed by observations they made as part of the authors’ direct participation in AI-related policy conversations in Australia and multilateral settings, not least the expert discussion on lethal autonomous weapons systems under the auspices of the United Nations Convention on Certain Conventional Weapons and the Responsible AI in the Military Domain (REAIM) Summits.

The paper is structured as follows. Section 1 discusses the importance of procurement for the responsible adoption of AI in the military domain, and why states’ urge to adapt procurement processes to accelerate the acquisition of military capability represents an important window of opportunity for implementing commitments and obligations around responsible military AI. Section 2 zooms in on the Australian context and maps out the country’s policy priorities on military AI. Section 3 offers recommendations on how Australia could reform its procurement process to balance the need for speed with the need for responsibility. 

Procurement and the Responsible Development and Use of AI in the Military Domain

Responsible AI in the Military Domain: from Principle to Practice

States widely regard AI as critical for the future of their military deterrence and warfighting capability. The reported use of AI in recent conflicts, and most recently the US’s and Israel’s war with Iran, has only reinforced that. For states that view military AI as critical, the key question is no longer if they should use AI for military purposes but how: what they should adopt, when, and how quickly.

As a recent report published by the UN Secretary-General noted, states broadly agree that rushing to adoption entails risks and that they should therefore seek to adopt AI responsibly—that is, with due regard for legal obligations, safety and security, as well as ethical considerations. This has led to the emergence over the past few years of multiple policy initiatives at the national and multilateral levels related to responsible AI. France, the US, the UK, Japan, Singapore and Australia have developed national principles on the development and use of military AI. NATO and the EU (via the European Parliament) have also set out principles to guide their member states. The Netherlands and the US have initiated two separate multilateral forums for states to discuss international principles––the REAIM Summit, and the US-led Political Declaration on Responsible Military Use of AI and Autonomy. These processes have led to the formulation of policy principles and commitments that many states around the world subscribe to. For instance, the Blueprint for Action adopted at the REAIM Summit in 2024 was endorsed by 60 states. The principles and policy commitments that came out of these various initiatives vary in formulation and purpose, but these commitments commonly touch on a core set of themes: (1) equitability and bias mitigation, (2) traceability and explainability, (3) reliability and security, (4) accountability and (5) governability.[7]

 A key task facing states now is to elaborate on how these very general principles are to be implemented in practice. This question was the core focus of the February 2026 REAIM Summit. A recurring talking point at the summit was that many of the principles require action in the design and development of AI systems. Surprisingly, little was said about the role of procurement, although it is one of the mechanisms through which these actions can be put into motion.

Procurement as a Key Intervention Point

Military procurement’s primary purpose is to enable capability adoption by bridging a military’s strategic needs and its operational capabilities. The strategic and security imperatives, as well as the complexity of some military capabilities, mean that many countries have introduced specific procurement processes dedicated to the military domain.[8] Military procurement is a formal and often sequential process that typically involves identifying a capability gap, formulating a technological need, soliciting and evaluating proposals from suppliers, and awarding and managing contracts, as well as testing and delivering capabilities to the armed forces.

In practice, military procurement processes are more than a process of acquiring capabilities; they also function as mechanisms for states to implement policy commitments and national and international legal obligations, including regarding responsible military AI.

Procurement has barely been discussed in policy debates on the responsible adoption of military AI, although it is an obvious vehicle for states to implement their commitments in this domain.[9] The procurement process provides many intervention points, from assessment of operational requirement to technology evaluation, to delivery management and review. It also provides a framework for states to think holistically about how their various commitments around responsibility can be tied together. These include their commitments to make military AI by design safe, secure, ethical and lawful; conduct legal reviews; conduct robust testing and evaluation; and ensure that end users are trained to understand how a capability may and may not be used.

Procurement Reform as a Window of Opportunity

The fact that procurement has not received much attention in debates on responsible military AI is all the more surprising as many states, the US in particular, have been vocal over the past decade about their intention to reform their military procurement processes to facilitate the adoption of AI and emerging technologies.[10] The common rationale for such reforms is threefold.

First, the pace of the procurement process is too slow. Procurement processes can take years, if not decades, largely due to an extensive bureaucratic decision-making process. The pace and model of military procurement stand in sharp contrast with the way in which and the speed at which AI products are typically developed and deployed in the civilian sector. AI product development cycles are measured in weeks or months.

Second, traditional procurement typically follows a waterfall model, moving sequentially through formal stages.[11] This linear approach assumes that capability requirements remain static throughout. By contrast, the development of AI capabilities—whether in the military or civilian domain—requires constant iteration; otherwise there is a risk that the technology once deployed is already obsolete. The inadequacy of waterfall methodologies for software systems was identified as early as the 1980s, and procurement agencies have been working to adapt their process accordingly, but with mixed success. It is not uncommon that military equipment enters into service while relying on software that is already outdated.

Third, military procurement is highly system/capability-centric: it is structured around the procurement of a specific weapon (e.g. type of missile), platform (e.g. aircraft) or system (e.g. air defence system). Critical AI capabilities may, however, operate as part of a system-of-systems involving various types of sensors, computing and communication infrastructure, and actuators (whether weapons or other military systems). This adds complexity, as it requires procurement agencies to think more holistically and consider many forms of technical and socio-technical interdependencies when articulating requirements for the design of the AI capability and the assurance process.

Fourth, the process and methods that procurement agencies typically use to verify that the technology they procure is reliable, safe and secure are not adequate for the latest advances of AI.[12] These assurance methods were initially designed for hardware and software that exhibit deterministic behaviour—that is, given the same inputs and conditions, they produce predictable, measurable, replicable and verifiable outputs.[13] AI capabilities, which are now based on machine learning, are, in contrast, non-deterministic. Formal methods of verification do not (yet) work well with them. The reliability, safety, and security of machine learning systems need to be evaluated through other means such as simulations and operational testing. Such empirical tests and evaluations are useful but can only provide a partial picture of how systems perform and how they may fail.

2. The Australian Context: Clear Focus on Speed with Less Guidance on Responsibility

Converging developments render the topic of military AI procurement timely and significant in the Australian context, where accelerating military acquisition of AI is on the agenda but there are relatively few (openly available) guidelines about responsible development and use of military AI. Relevant developments include Australia’s participation in AUKUS Pillar II, Australia’s engagement in multilateral processes on responsible AI and, more recently, the publication of the Department of Defence’s Policy Settings for Responsible Use of Artificial Intelligence in Defence.

The Need for Speed: AUKUS Pillar II and the Advanced Strategic Capabilities Accelerator

Australia’s participation in AUKUS Pillar II and its Advanced Strategic Capabilities Accelerator (ASCA) reflects a clear strategic orientation towards accelerated military AI adoption, particularly in the development and fielding of AI-enabled capabilities and autonomous systems.

AUKUS Pillar II addresses six technological areas, including AI and autonomy.[14] A working group is dedicated to each technological area, as well as to two functional areas: innovation (accelerating the countries’ respective defence innovation enterprises ‘to more rapidly integrate commercial technologies to solve warfighting needs’) and information sharing to enable work related to Pillar II.[15] Working groups may engage in a number of activities to accomplish their goals,[16] and much AUKUS-related activity relating to advanced capabilities is classified.[17] Despite this, it is clear that compressing development and acquisition timelines is an ambition of Pillar II. A 2022 fact sheet described early work in the ‘AI and autonomy’ working group as being focused on ‘accelerating adoption, and improving the resilience of, autonomous and AI-enabled systems in contested environments’.[18] The three AUKUS states’ defence ministers announced in 2024 that ‘the National Armaments Directors from each nation are identifying opportunities to harmonise acquisition processes and reducing barriers to facilitate the accelerated delivery of Pillar II advanced capabilities’.[19]

Established in July 2023 following the Defence Strategic Review, ASCA functions as a dedicated mechanism for accelerating the transition of advanced capabilities from development into operational use.[20] ASCA’s stated mission is to accelerate the development and transition of asymmetric capabilities to the ADF through innovation, in order to meet Defence priority needs.[21] ASCA does not develop capabilities itself; rather, it connects and streamlines the defence innovation system to drive capability development and acquisition pathways at speed,[22] coalescing with Australia’s longstanding concern about sovereign defence industrial capability.[23] ASCA is pursuing time-limited ‘missions’ that respond to priorities with minimum viable capabilities.[24] A priority for the program is ‘trusted autonomy’,[25] and these discrete missions encompass AI capabilities, such as ‘Ghost Shark—Mission Zero’, which is an underwater autonomous vehicle program,[26] and ‘Mission Syracuse’ to support a counter small uncrewed aerial system capability.[27] Alongside these discrete missions, ASCA has established a dedicated program for longer-horizon AI development. The Emerging and Disruptive Technologies (EDT) Decision Advantage program is oriented towards ‘discovery, new knowledge, concepts and prototypes’ in areas including automated data integration, machine reasoning, AI agents and AI orchestration.[28]

The Need for Responsibility: International Commitments and National Policy

When it comes to the need for responsibility, Australia’s orientation has remained more general. Australia has been an active participant in the principal international forums on responsible military AI. It participated in the three REAIM summits—at The Hague in 2023, Seoul in 2024 and A Coruña in 2026—and has endorsed the outcome documents of each summit.[29] Australia has also endorsed the US-led Political Declaration on Responsible Military Use of AI and Autonomy (the Political Declaration)—a framework launched in 2023 to build international consensus around responsible behaviour and guide states’ development and use of military AI.[30] In its submission to the UN Secretary-General’s report on AI in the military domain, Australia affirmed that the military use of AI should be ethical, responsible, and enhance international security, and must comply with applicable international law, including international humanitarian law (IHL) and international human rights law, throughout the capability lifecycle.[31]

Yet at the national level, Australia’s measures to give effect to these obligations and commitments are less defined. Australia has published only one policy document on responsible AI in the military domain (others may exist but are not available publicly). The Policy Settings for Responsible Use of Artificial Intelligence in Defence, published in March 2026, are intended to implement Australia’s international commitments and its obligations under international law. The settings include three broad requirements relating to the use of AI in the Australian Defence Force: lawfulness, adherence to values-based principles, and proportionate controls.[32] The document outlines control measures, such as testing and evaluation, post-incident/impact reviews, and reporting and oversight; and control models, such as direct human oversight and supervision, human–machine teaming and machine–machine teaming, that it says will be applied based on levels of risk and the characteristics of specific AI technologies.[33] The document is intended to fill a carve-out for the Defence portfolio in the Australian Government’s 2024 Policy for the Responsible Use of AI in Government but is considerably less detailed than this antecedent.[34]

The result is that Australia currently occupies an awkward intermediate position. It has made important commitments to responsible military AI and reaffirmed relevant obligations under international law. But it has not articulated the institutional architecture for translating these commitments into practice, including in the procurement phase.

How Australia manages this tension will be consequential for its relationship with partner states like the UK and the US. The UK has more elaborate national policies on responsible military AI, including the 2022 Defence AI Strategy and the 2024 Directive on Dependable AI in Defence.[35] The US has similar—if more detailed—policies, developed even earlier, including the Department of Defense Directive 3000.09, ethical principles for AI, and Responsible AI Strategy and Implementation Pathway.[36] But in January 2026, the US announced its AI strategy for the (now) Department of War. This strategy shifted emphasis to speed and lethality, mandating ‘any lawful use’ of AI and directing removal of barriers to rapid deployment. Other developments reinforce the uncertainty about the US position with respect to responsible military AI and the signal it sends to allies and partners. This includes the absence of progress in the process led by the US Department of State to elaborate implementation measures under the Political Declaration. Moreover, a dispute with Anthropic, a supplier, over the permissible use of its AI model by the US military saw the Department of War threaten to terminate the company’s contract and designate it a ‘supply-chain risk’ after Anthropic refused to allow its Claude model to be used for fully autonomous lethal targeting and domestic mass surveillance.[37] 

This tension between strategic orientation towards rapid procurement of military AI and high-level commitments to responsible military AI is also consequential for the significant reforms Australia is undertaking of its defence procurement architecture. This moment offers an opportunity for Australia to design and put in place the institutional architecture and processes necessary to give effect to its commitments around responsible AI, thereby developing a national approach and clarifying its position in relation to its allies. The next section offers a few recommendations to that effect. 

3. Seizing the Opportunity to Embed Responsible Military AI in a New Procurement Environment

In 2025, Australia announced significant defence procurement reforms, including the creation of a Defence Delivery Agency (DDA) that consolidates the functions and budgets of Australia’s principal procurement organisations—the Capability Acquisition and Sustainment Group), the Guided Weapons and Explosive Ordnance Enterprise and the Naval Shipbuilding and Sustainment Group (though notably not the Australian Submarine Agency).[38] The drivers for this reform are familiar: long delays, cost overruns and blurred lines of responsibility.[39] The establishment of the DDA reflects a recognition that existing acquisition processes do not deliver capability at the pace, scale and price that Australia demands. With respect to technologies such as military AI, the 2024 National Defence Strategy stated that Australia is ‘reforming its approach to contracting with defence industry’ and developing ‘[m]ore flexible contracting mechanisms … to support agile and developmental projects where technology is rapidly changing or where the ADF’s requirements cannot be fully known’.[40]

This moment offers an opportunity to implement legal obligations and operationalise high-level policy commitments to responsible development and use of military AI through procurement processes.

Procurement can serve as a mechanism for implementing responsible military AI, but only if deliberately structured to do so. Procurement represents a critical juncture, prior to the use of military AI capabilities, where legal obligations and high-level policy commitments can be operationalised. A confluence of factors—sharpened focus on accelerated adoption through AUKUS Pillar II and ASCA, new responsible military AI policy settings and structural procurement reform—creates a policy opportunity for Australia. The following recommendations are aimed at ensuring that the institutional architecture for procuring military AI technologies at pace is built with responsible procurement in mind.

Connect Responsible Development and Use of Military AI to Responsible Procurement

A critical first step in giving effect to legal obligations and policy commitments regarding responsible military AI is to ensure that policies, procedures and practices for military procurement explicitly refer to any national laws, policies and commitments related to responsible development and use of military AI.[41] Without explicit cross-referencing, responsible AI principles risk remaining siloed within governance documents that are institutionally distant from procurement decision-makers.

For Australia, each component of the suite of new policy and procedural documents that will necessarily accompany the establishment of the DDA—including its founding instruments, manuals, templates and oversight frameworks—should make explicit reference to Australia’s commitments under IHL, Defence’s policy settings, and outcome documents of REAIM summits and the Group of Governmental Experts on Lethal Autonomous Weapons Systems, as well as the US-led Political Declaration. Similarly, the terms of reference and operational procedures being developed for ASCA missions involving AI-enabled capabilities should be reviewed against these commitments. Australia’s policy settings for responsible military AI set the example, expressly referencing Australia’s commitments to REAIM outcome documents, the Political Declaration and international law.

Independent Capacity to Test Supplier Claims Within the DDA

Australia should ensure that the design of the DDA explicitly includes independent testing, evaluation, validation and verification (TEVV) capacity for AI-enabled capabilities, whether in-house or using trusted third-party providers. Legal obligations and policy commitments regarding responsible military AI require procurement authorities to maintain independent capacity to test supplier claims.[42] Assessing supplier claims requires the technical literacy to interrogate whether and why a military AI capability is needed, and to evaluate what suppliers assert about the performance and reliability of their systems. And the pace at which Australia seeks to transition from prototyping to fielding AI capabilities makes it even more important to ensure that the DDA is designed and resourced at the outset with the technical expertise and institutional authority to conduct or commission independent TEVV rather than relying on supplier-provided assurance.

Integrate Legal Review Requirements into ASCA and DDA Processes

ASCA missions and DDA processes involving AI-enabled capabilities involve precisely the kinds of activities that trigger the need for a legal review under Article 36 of Additional Protocol I to the Geneva Conventions. The obligation to conduct such legal reviews is addressed prominently in Australia’s policy settings for responsible military AI, reflecting its longstanding focus on the importance of such reviews in the governance of autonomous weapon systems and other applications of AI in the military domain.[43] Legal review requirements should be explicitly embedded within ASCA’s mission governance framework and within the DDA acquisition procedures.

In doing so, Australia should address the challenges of reviewing military AI capabilities. The integration of AI into military systems introduces distinctive characteristics that challenge traditional legal review processes and methodologies.[44] Critically, the scope of Article 36 is limited to the review of new weapons, means and methods of warfare.[45] This raises interpretative questions about which military AI capabilities to review and when, which could be clarified in new procedures.[46] For example, even where AI capabilities are subject to legal reviews, the iterative and fast-paced approach taken to the acquisition of some AI-enabled military capabilities, and the challenges of reliability and assurance associated with opaque and brittle systems, potentially complicate the linear process of legal reviews within the procurement process.[47] Reform of procurement offers an opportunity to adapt legal review processes to these challenges, including by subjecting all military AI capabilities to legal review, regardless of whether they fall within the Article 36 category of a new weapon, means or method of warfare.

Develop and Publish Guidance for Industry on Responsible Military AI

Australia should develop and publish documents articulating clear expectations for suppliers of military AI capabilities, addressing the technical parameters and performance standards necessary to implement its policy settings, including the values of accountability; bias and harm mitigation; explainability; human-centrism; and reliability and security.[48]

Australia’s approach to military AI procurement—including ASCA and the DDA—draws in a wide range of industry actors, including Australian primes, small and medium-sized businesses, and research organisations. Many of these actors will not have established internal frameworks for translating responsible military AI principles into product specifications. Without such guidance, procurement authorities and suppliers alike face a more complex and lengthier process of pre-contract evaluation, back-and-forth requests for documentation, and uncertainty around requirements, which can put at risk the delivery of the capability on time and on budget.[49] The implementation measures directed at giving effect to Defence’s 2026 policy settings provide a moment to initiate the development of industry-facing guidance.

Take Opportunities to Broaden and Strengthen Responsible Procurement of Military AI with Partners and Neighbours

States adapting their procurement processes are navigating fundamental implementation questions independently, without the benefit of shared learning or common vocabularies. Responsible procurement should be an explicit component of international policy discussions on military AI governance.[50] Opportunities for such conversations exist at the REAIM summits, at the ongoing workshops among states supporting the US-led Political Declaration and the US-led AI Partnership for Defence, and at any discussion on AI in the military domain under the auspices of the UN General Assembly First Committee.[51] Australia is well positioned to contribute to addressing this gap, both within its principal security partnership and across its broader region.

Within AUKUS, the existing commitment to harmonising capability development and acquisition systems creates an opportunity for developing common responsible AI procurement standards among Australia, the US and the UK. Australia should advocate for a responsible AI procurement component within that harmonisation agenda, such as shared TEVV frameworks and common supplier expectations across the three partners.

Australia also maintains defence relationships with a range of regional partners such as Indonesia, Japan, the Philippines, the Republic of Korea, Singapore and Thailand, who are themselves beginning to procure AI-enabled military capabilities.[52] Australia is well placed to work with regional partners to discuss the development of governance frameworks and strengthening of procurement institutions.[53]

Conclusion

The rapid adoption of AI in the military domain should not come at the expense of responsibility. Procurement is a juncture at which policy commitments and legal obligations can be given practical effect. Australia has made commitments to responsible military AI, and is accelerating adoption through AUKUS Pillar II and ASCA. Forthcoming procurement reforms in Defence provide the opportunity to connect commitments and obligations to actual practice.

The recommendations set out in this paper are aimed at ensuring that new procurement architecture is built with responsibility in mind from the outset. Embedding responsible AI principles at this stage will avoid the need for retrofitting them—a slower, costlier and less effective process. Beyond the national level, Australia has an opportunity to shape how responsible procurement is understood among AUKUS partners and across the region—contributing not only to its own practice but to the emerging frameworks for international governance of military AI.

Endnotes

[1] Australian Department of Defence, 2026 National Defence Strategy (Canberra: Commonwealth of Australia, 2026), p. 80.

[2] See for example discussion in Australian Department of Defence, 2026 National Defence Strategy, p.16.

[3] ‘Remarks by Deputy Secretary Work on Third Offset Strategy’, U.S. Department of Defense (website), 28 April 2016, at: www.war.gov/News/Speeches/Speech/Article/753482/remarks-by-deputy-secretary-work-on-third-offset-strategy.

[4] S Shaikh and W Rumbaugh, ‘The Air and Missile War in Nagorno-Karabakh: Lessons for the Future of Strike and Defense’, Center for Strategic and International Studies (CSIS) (website), 8 December 2020; K Bondar, ‘Ukraine’s Future Vision and Current Capabilities for Waging AI-Enabled Autonomous Warfare’, CSIS, 20 March 2025; B McKernan and H Davies, ‘The Machine Did It Coldly’: Israel Used AI to Identify 37 000 Hamas Targets’, The Guardian, 4 April 2024; R Booth and D Milmo, ‘Iran War Heralds Era of AI-Powered Bombing Quicker than ‘Speed of Thought’, The Guardian, 3 March 2026. See also Ingvild Bode and Tom FA Watts, Loitering Munitions and Unpredictability: Autonomy in Weapon Systems and Challenges to Human Control (Odense: Centre for War Studies, University of Southern Denmark, 2023), pp. 52–66, at: https://findresearcher.sdu.dk/ws/portalfiles/portal/231643063/Loitering_Munitions_Unpredictability_WEB.pdf.

[5] For example, several NATO allies formalised visions for AI adoption in national strategies for their militaries, including France, the UK, Canada and the USA. See French Ministry of Armaments, L’Intelligence Artificielle au Service de la Défense [Artificial Intelligence in Support of Defence], Report of the AI Task Force (Ministère des Armes, 2019); British Ministry of Defence (MOD), Defence Artificial Intelligence Strategy (London: MOD, June 2022); Canadian Department of National Defence (DND), The Department of National Defence and Canadian Armed Forces Artificial Intelligence Strategy (Ottawa: DND, 2024); US Secretary of War, ‘Accelerating America’s Military AI Dominance’, memorandum, 9 January 2026.

[6] On state narratives, see for example ‘Putin: Leader in Artificial Intelligence Will Rule World’, AP News, 2 September 2017; U.S. Department of Defense (US DOD), 2018 Department of Defense Artificial Intelligence Strategy: Harnessing AI to Advance Our Security and Prosperity (US DOD, February 2019). See also V Boulanin et al., Artificial Intelligence, Strategic Stability and Nuclear Risk (Stockholm: SIPRI, 2020). On the role of industry, see for example S Ghaffray, ‘OpenAI Emphasizes China Competition in Pitch to a New Washington’, Financial Post, 13 January 2025; M Sigalos, ‘OpenAI’s Altman Warns the US Is Underestimating China’s Next-Gen AI Threat’, CNBC, 18 August 2025; S Elbaum and J Panter, ‘DOD’s AI Balancing Act’, Council on Foreign Relations (website), 2 December 2025; J O’Donnell, ‘OpenAI’s New Defense Contract Completes Its Military Pivot’, MIT Technology Review, 4 December 2024.

[7] Netta Goussac and Vincent Boulanin, Responsible Procurement of Military Artificial Intelligence (Stockholm: SIPRI, February 2026), pp. 17–18, at: https://doi.org/10.55163/YOLG1827.

[8] OECD, Brief 23: Public Procurement—Defence Procurement (OECD, 2011), p. 3, at: www.oecd.org/content/dam/oecd/en/publications/reports/2011/09/defence-procurement_g17a25f1/5js4vmntfz0p-en.pdf.

[9] An early examination of the issue appears in S Allik et al., ‘A Framework for Human Decision-Making Through the Lifecycle of Autonomous and Intelligent Systems in Defense Applications’, in A Framework for Human Decision-Making Through the Lifecycle of Autonomous and Intelligent Systems in Defense Applications (IEEE SA, 2024), pp. 42–44.

[10] See for example G Gentile et al., A History of the Third Offset, 2014–2018, RAND Research Report RR-A454-1 (RAND, 2021); K Hicks, US Deputy Secretary of Defense, ‘The Urgency to Innovate’, keynote address, NDIA Emerging Technologies for Defense Conference, Washington DC, 28 August 2023.

[11] J Wallin, ‘Safe and Effective: Advancing Department of Defense Test and Evaluation for AI and Autonomous Systems’, Center for New American Security (website), 13 March 2025; JS Gansler, W Lucyshyn and A Spiers, Using Spiral Development to Reduce Acquisition Cycle Times (University of Maryland Center for Public Policy and Private Enterprise, 2008).

[12] C Urban and A Miné, ‘A Review of Formal Methods Applied to Machine Learning’, arXiv, 21 April 2021.

[13] RS Panwar, Q Li and JNT Shanahan (eds), Military Artificial Intelligence Test and Evaluation Model Practices (INHR, 2024); Wallin, ‘Safe and Effective’.

[14] Government of the United Kingdom, Fact Sheet: Implementation of the Australia–United Kingdom–United States Partnership (AUKUS) (Prime Minister’s Office, 2022), at: www.gov.uk/government/publications/implementation-of-the-australia-united-kingdom-united-states-partnership-aukus.

[15] Ibid.

[16] Congressional Research Service, AUKUS Pillar 2 (Advanced Capabilities): Background and Issues for Congress, R47599 (Washington DC: Congressional Research Service, May 2024), p. 4, at: www.congress.gov/crs-product/R47599.

[17] U.S. Department of Defense, ‘AUKUS Defense Ministers Meeting Joint Statement’, press release, 1 December 2023, at: www.defense.gov/News/Releases/Release/Article/3604511/aukus-defense-ministers-meeting-joint-statement.

[18] Government of the United Kingdom, Fact Sheet: Implementation of the Australia–United Kingdom–United States Partnership (AUKUS).

[19] U.S. Department of Defense, ‘AUKUS Defence Ministers’ Meeting Communiqué’, press release, 26 September 2024, at: www.defense.gov/News/Releases/Release/Article/3918402/aukus-defence-ministers-meeting-communique.

[20] Australian Department of Defence, ‘Government Announces Most Significant Reshaping of Defence Innovation in Decades to Boost National Security’, media release, 28 April 2023, at www.minister.defence.gov.au/media-releases/2023-04-28/government-announces-most-significant-reshaping-defence-innovation-decades-boost-national-security.

[21] Ibid.

[22] ‘Advanced Strategic Capabilities Accelerator’, Australian Government Directory, at: www.directory.gov.au/portfolios/defence/department-defence/vice-chief-defence-force/advanced-strategic-capabilities-accelerator (accessed 23 March 2026).

[23] Tanya Monro, ‘How Can Australia Build Sovereign Capability in AI, and Why Is It That Important?’, Australian Institute for Machine Learning, University of Adelaide (website), 28 July 2022, at www.adelaide.edu.au/aiml/news/list/2022/07/28/how-can-australia-build-sovereign-capability-in-ai-and-why-is-it-that.

[24] ‘Missions’, Advanced Strategic Capabilities Accelerator (website), at: www.asca.gov.au/activities/missions (accessed 23 March 2026).

[25] Australian Department of Defence, ‘Government Announces Most Significant Reshaping of Defence Innovation in Decades to Boost National Security’.

[26] ‘Ghost Shark—Mission Zero’, Advanced Strategic Capabilities Accelerator (website), at: www.asca.gov.au/activities/missions/ghost-shark-mission-zero (accessed 23 March 2026).

[27] ‘Mission Syracuse’, Advanced Strategic Capabilities Accelerator (website), at: www.asca.gov.au/activities/missions/mission-syracuse (accessed 23 March 2026).

[28] ‘Emerging and Disruptive Technologies: Decision Advantage’, Advanced Strategic Capabilities Accelerator (website), at: www.asca.gov.au/activities/emerging-disruptive-technologies/decision-advantage (accessed 23 March 2026); Australian Department of Defence, ‘Harnessing Tech to Drive Faster Decisions across Defence’, news release, 5 January 2026, at: www.defence.gov.au/news-events/releases/2026-01-05/harnessing-tech-drive-faster-decisions-across-defence.

[29] REAIM 2023 Call to Action adopted 16 February 2023 (Government of the Netherlands, 2023); REAIM Blueprint for Action adopted 10 September 2024 (Government of the Republic of Korea, 2024); REAIM 2026 Pathways to Action adopted 5 February 2026 (Government of Spain, 2026).

[30] U.S. Department of State, Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy (Bureau of Arms Control, Deterrence and Stability, 2023), at: www.state.gov/bureau-of-arms-control-deterrence-and-stability/political-declaration-on-responsible-military-use-of-artificial-intelligence-and-autonomy.

[31] Australian Government, ‘Australia’s Submission to the United Nations Secretary-General’s Report on Artificial Intelligence in the Military Domain’, response to UN General Assembly Resolution 79/239, 2025, at: https://docs-library.unoda.org/General_Assembly_First_Committee_-Eightieth_session_(2025)/79-239-Australia-en.pdf.

[32] Australian Department of Defence, Policy Settings for Responsible Use of Artificial Intelligence in Defence (Canberra: Commonwealth of Australia, March 2026), at: www.defence.gov.au/sites/default/files/2026-03/Policy-Settings-for-Responsible-Use-of-Artificial-Intelligence-in-Defence-%5BOFFICIAL%5D.pdf.

[33] Ibid, p. 9.

[34] Digital Transformation Agency Policy for the Responsible Use of AI in Government, Version 2.0 (Commonwealth of Australia, 2025).

[35] MOD, Defence Artificial Intelligence Strategy; British Ministry of Defence, Dependable Artificial Intelligence (AI) in Defence—Part 1: Directive, JSP 936 V1.1 (MOD, 2024).

[36] U.S. Department of Defense, Autonomy in Weapon Systems, Directive 3000.09 (US DOD 2023); U.S. Department of Defense, ‘DOD Adopts Ethical Principles for Artificial Intelligence’, press release, 24 February 2020; U.S. Department of Defense, Responsible AI Working Council, U.S. Department of Defense Responsible Artificial Intelligence Strategy and Implementation Pathway (Washington DC: US DOD, 2022).

[37] See Jessica Dorsey et al., ‘The Pentagon/Anthropic Clash Over Military AI Guardrails’, Opinio Juris, 26 February 2025; Congressional Research Service, Pentagon–Anthropic Dispute over Autonomous Weapon Systems: Potential Issues for Congress, CRS Insight IN12669 (Washington DC: Congressional Research Service, 2026), at: www.congress.gov/crs-product/IN12669. At the time of writing, the dispute remained unresolved and before the courts.

[38] Richard Marles and Pat Conroy, ‘Reforming Defence Capability Development and Delivery’, media release, 1 December 2025, at: www.minister.defence.gov.au/media-releases/2025-12-01/reforming-defence-capability-development-delivery.

[39] Australian Department of Defence, 2023 Defence Strategic Review (Canberra: Commonwealth of Australia, 2023), para. 12.5, at: www.defence.gov.au/about/reviews-inquiries/defence-strategic-review.

[40] Australian Department of Defence, 2024 National Defence Strategy (Canberra: Commonwealth of Australia, 2024), para. 8.4, at: www.defence.gov.au/about/strategic-planning/2024-national-defence-strategy-2024-integrated-investment-program.

[41] Goussac and Boulanin, Responsible Procurement of Military Artificial Intelligence, p. 25.

[42] Ibid., p. 19.

[43] Australian Government, The Australian Article 36 Review Process, working paper submitted to the Group of Governmental Experts on Lethal Autonomous Weapons Systems, Convention on Certain Conventional Weapons, August 2018, at: https://docs-library.unoda.org/Convention_on_Certain_Conventional_Weapons_-_Group_of_Governmental_Experts_(2018)/2018_GGE%2BLAWS_August_Working%2Bpaper_Australia.pdf.

[44] V Boulanin and M Verbruggen, Article 36 Reviews: Dealing with the Challenges Posed by Emerging Technologies (Stockholm: SIPRI, 2017); N Goussac and R Liivoja, ‘Legal Review of Military Artificial Intelligence Capabilities’, Articles of War, 25 August 2025; T Vestner and I Rossi, ‘Legal Reviews of War Algorithms’, International Law Studies 97, no. 509 (2021): 512.

[45] Article 36, Protocol Additional to the 1949 Geneva Conventions, and Relating to the Protection of Victims of International Armed Conflicts (API), opened for signature 12 December 1977, entered into force 7 December 1978; International Committee of the Red Cross (ICRC), A Guide to the Legal Review of New Weapons, Means and Methods of Warfare (Geneva: ICRC: 2006).

[46] Goussac and Liivoja, ‘Legal Review of Military Artificial Intelligence Capabilities’.

[47] Ibid.

[48] Australian Department of Defence, Policy Settings for Responsible Use of Artificial Intelligence in Defence, p. 8.

[49] Goussac and Boulanin, Responsible Procurement of Military Artificial Intelligence.

[50] Ibid.

[51] On the AI Partnership for Defence, see US DOD, Chief Digital and Artificial Intelligence Office, ‘DOD CDAO Holds Eleventh AI Partnership for Defense’, press release, 13 August 2025. On discussions under the auspices of the UN General Assembly, see United Nations General Assembly, First Committee, Artificial Intelligence in the Military Domain and Its Implications for International Peace and Security, A/C.1/80/L.46 (UN, 2025), para. 10.

[52] Gatra Priyandita, ‘When Regulating AI, ASEAN Should Remain Committed to Responsible Behaviour’, Fulcrum, 4 June 2024, at: https://fulcrum.sg/aseanfocus/when-regulating-ai-asean-should-remain-committed-to-responsible-behaviour; Kendall Stiles, Asfandyar Mir, Jana Puglierin and Karisa Pevehouse, ‘Editorial’, Global Policy 13, no. 3 (2022): 319–321, at: https://doi.org/10.1111/1758-5899.13106.

[53] Japan, the Republic of Korea and Singapore have developed their own responsible military AI frameworks: see Japanese Ministry of Defense, Guideline for Responsible AI Application in Research and Development of AI-Equipped Defense Systems, Version 1 (provisional translation) (Ministry of Defense, June 2025); Suon Choi, South Korean Military-AI Integration: Opportunities and Risks (Seoul: Asia-Pacific Leadership Network, 2026), at: https://cms.apln.network/wp-content/uploads/2026/03/Choi_March_2026.pdf; Ng Eng Hen, ‘Welcome Address at the 3rd Singapore Defence Technology Summit’, speech, 12 October 2021, at: www.mindef.gov.sg/news-and-events/latest-releases/12oct21_speech.