31 Jul Human rights in the shadow of military exceptionalism: reflections on the Informal Exchange on Artificial Intelligence in the military domain
[Chantal Joris is a Swiss qualified lawyer and the acting Head of Law and Policy at ARTICLE 19.
Dr Ilia Siatitsa is Programme Director at Privacy International.
Tomaso Falchetta is Global Advocacy Coordinator at Privacy International.]
On 15–17 June 2026, the United Nations Office for Disarmament Affairs (ODA) hosted the first Informal Exchange on Artificial intelligence in the military domain and its implications for international peace and security. The meeting marked the second stage of a UN General Assembly process initiated by Resolution 79/239 (2024), which requested the UN Secretary-General (UNSG) to gather views from states and other stakeholders on AI in the military domain and prepare a report. Building on that report, UNGA Resolution 80/58 (2025) mandated a three-day informal exchange in Geneva to share views on observations and conclusions of the UNSG’s report.
The process represents an important development. Until recently, the multilateral discussions of military AI which started fifteen years ago were largely confined to lethal autonomous weapon systems (LAWS) (see here for the 2026 draft final report and discussions). While the topic is urgent, LAWS ultimately constitute a narrow subset of the military uses of AI. In particular the role of AI decision support systems (AI DSS), which describe computerised tools designed to aid human decision-making, has received increased attention. The UN General Assembly rightly attempted with this exchange to broaden the conversation. Rather than replacing the discussion on LAWS, the intention was to complement it with a space to consider the implications of AI across the military domain and, importantly, throughout the technology’s lifecycle.
As news and analysis of the uses of AI across conflicts in Gaza, Ukraine, Venezuela and Iran have continued to proliferate, the Informal Exchange – which was open to all UN Member and observer States – as well as international and regional organisations, the International Committee of the Red Cross (ICRC), academia, civil society and industry – was attended with quite some anticipation. While there was a high level of participation from governments across regions, it is notable that some of the most prominent adopters and promoters of AI in the military domain, such as the United States of America and Israel, as well as most of the big tech companies operating in this sector, did not actively participate in the discussion.
The official discussions offered insight into the assumptions that continue to shape multilateral debates on the governance of military AI. Over the course of the three days and thirteen sessions (summarised here), states and other stakeholders, including civil society, experts and academics, moved from definitional questions over what “AI in the military domain” even means, through every stage of the technology’s life-cycle, into broader areas such as: AI and nuclear powers, cyber operations, biological and chemical weapons, risks of unintended escalation, as well as the capacity gap between advanced military powers and other states. By Wednesday’s close, more questions were open than closed, and no consensus had formed on where this conversation goes next, although there was general agreement on the need to have further exchanges.
During the Informal Exchange, three aspects stood out to us. First, during the official discussions, military AI continued to be treated as a distinct domain, largely separate from broader debates on AI governance. Second, the normative discussion centred overwhelmingly on international humanitarian law, with comparatively little attention paid to the role of international human rights law across the AI lifecycle. This was particularly evident in discussions of data, which was largely treated as an inevitable feature of military AI rather than as something whose collection and use require an independent legal basis and governance framework. Third, despite the central role private companies play in developing, supplying and maintaining military AI capabilities, they featured only marginally in the official debate.
These limitations in the official discussions stood in contrast to the broader range of conversations explored in side events organised on the margins of the Informal Exchange. It was important that this space existed – but it also points to the need to include these issues and the stakeholders who raise them in the formal debate.
The case against military exceptionalism
The Informal Exchange confirmed a persistent implicit view among many states that military applications of AI should be governed as a distinct domain if at all. Yet military AI does not exist in isolation. It is developed within the same technological ecosystem as civilian AI, frequently builds on general-purpose models and commercially developed technologies, and increasingly supplied by companies selling into both civilian and military markets. Treating military AI as entirely separate from broader AI governance is therefore becoming increasingly difficult to justify.
As Giacomo Persi Paoli, Head of Security and Technology Programme of the United Nations Institute for Disarmament Research (UNIDIR), pointed out during the single hour dedicated to “Existing and Emerging Normative Proposals”, rather than starting from scratch, discussions on military AI should build on the governance frameworks that states have already developed for AI more broadly. There is no shortage of ongoing initiatives, frameworks and standards as well as legally binding treaties and UN General Assembly resolutions. Over the past decade, they have developed an extensive body of AI governance frameworks, standards and legal instruments, including the UNESCO Recommendation on the Ethics of Artificial Intelligence, the Global Digital Compact, the Council of Europe Framework Convention on Artificial Intelligence, and the EU AI Act. While LAWS discussions started early, debates on military AI more broadly have lagged behind their civilian AI counterparts.
Across this body of AI governance, states have already widely accepted baseline principles on transparency, accountability, oversight and the relevance for human rights for AI generally. Further, the UN General Assembly already called on states in Resolution 78/265 (2024) to refrain from using AI systems that are impossible to operate in compliance with international human rights law or that pose undue risks to the enjoyment of human rights. Military AI should build on those foundations, which remain applicable and could be complemented by tailored rules where the specificities of the military domain genuinely require it. The same exceptionalism also influences which legal frameworks are brought to bear on military AI. Across AI governance more broadly, international human rights law has become a central point of reference. In the military context, however, it remained largely at the margins of the discussion.
The limits of international humanitarian law as the dominant framework
Despite the broader mandate of Resolution 80/58 (2025), much of the legal discussion in the official sessions of the Informal Exchange was framed predominantly around the – positive or negative – impact of the use of AI systems on conflict parties’ ability to comply with international humanitarian law, including the principles of distinction, proportionality, and precaution, as well as on concepts such as “meaningful human control,” “human in the loop,” and the risks posed by AI bias and system errors. These are essential questions. They are not, however, the only ones that need to be addressed.
Most delegations affirmed what Resolution 80/58 (2025) also recognises from the outset that ‘international law, including the Charter of the United Nations, international humanitarian law and international human rights law,’ applies to the use of AI in the military domain. Yet, while human rights were explicitly mentioned a number of times, it was far less engaged with substantively. Rights such as privacy and freedom of expression were mentioned only in passing, rather than examined as legal constraints that should shape the development and deployment of military AI.
This was particularly evident in discussions of data. Military AI systems routinely rely on datasets containing biometric information, facial-recognition outputs, and patterns of movement and communication derived from mass surveillance. Yet these datasets were largely treated as neutral technical inputs and an inevitable feature of military AI, rather than as the product of data collection and processing that require a legal basis and must comply with human rights principles of legality, legitimacy, necessity and proportionality.
The relative absence of references to international human rights law is perhaps unsurprising given the institutional setting. Convened by the UN Office for Disarmament Affairs (ODA), the exchanges in the official sessions naturally centred on weapons and international humanitarian law, as most of the invited discussants were humanitarian law specialists. Instead, the relevance of international human rights law throughout the entire life cycle of AI systems was repeatedly underlined during some of the side events, with panellists including representatives of the Office of High Commissioner for Human Rights and civil society organisations highlighting the need to bridge those gaps.
In official settings, however, human rights obligations are mostly addressed as ethical considerations rather than binding legal requirements. This is an important point of departure. Military AI has become the subject of numerous initiatives on “responsible” and “ethical” use and there is a real risk that governance discussions lose sight of the fact that military applications of AI are already subject to binding legal obligations under international human rights law. Ethical principles can usually inform the development and use of these technologies, but they cannot substitute for the application of law.
Setting to one side the broader debate about the precise relationship between international human rights and humanitarian law, it is clear that the former, including the rights to privacy and free expression, continue to apply during armed conflict. What is more, especially for military AI, major decisions across the technology’s life cycle – research, procurement, everyday use – fall squarely within the ordinary scope of international human rights law. The same is true of data collection and processing: mass surveillance conducted to build training datasets, or biometric registration used to populate target lists, engages the rights to privacy and non-discrimination regardless of whether the resulting system is ultimately used inside or outside an armed conflict.
What about the private sector?
A third notable omission from the official discussions was the role of corporate actors. Military AI is not developed or deployed by states alone – quite the contrary. Some of the world’s largest technology companies have become key suppliers to the military. Across the AI lifecycle, private companies provide the computing infrastructure, cloud services and specialised hardware needed to train and operate AI systems. They develop the foundation models that increasingly underpin military AI applications, build tailored decision-support systems, and supply the software, maintenance and technical expertise on which these systems depend. Military AI is therefore deeply embedded in the commercial AI ecosystem. Despite this, the Informal Exchange focused almost exclusively on states and their obligations under international law. It paid little attention to states’ obligations to regulate corporate conduct, including through human rights requirements in public procurement, and to corporate responsibilities to respect human rights under the UN Guiding Principles on Business and Human Rights (UNGPs).
This again stands in contrast to broader AI governance. For example, the EU AI Act imposes direct obligations on providers and deployers of certain AI systems. The Council of Europe Framework Convention on Artificial Intelligence, Human Rights, Democracy and the Rule of Law requires states parties to ensure that private actors operating within their jurisdiction or control respect the Convention’s principles, including with respect to human dignity, accountability, and privacy and personal data protection. Although a soft law instrument, the UNGPs have been highly influential and have been endorsed and incorporated by various UN bodies, including in discussions concerning private actors developing and deploying AI technologies. The UNGPs establish that companies have an independent responsibility to respect human rights by identifying, preventing, mitigating, and accounting for adverse human rights impacts arising throughout their operations and business relationships. They further affirm that companies must respect international humanitarian law. These responsibilities apply equally to companies that develop or supply AI systems for military purposes.
As military AI increasingly relies on commercially developed technologies, meaningful governance will require addressing corporate responsibility alongside state responsibility. Leaving the role of companies at the margins overlooks one of the principal actors shaping the development and deployment of military AI and risks creating significant gaps in accountability across the technology’s lifecycle. This message was emphasised during a side event on the weaponisation of data‑intensive technologies, with a particular focus on the role of Big Tech and the military‑industrial complex, organised by PAX and Privacy International, but was not reflected during the official debate.
A window of opportunity
The Informal Exchange marked an important and urgent step in broadening multilateral discussions beyond LAWS to consider AI across the military domain. It demonstrated growing international engagement with the issue and confirmed a broad consensus that international law, including the UN Charter, international humanitarian law, and international human rights law, provides the normative framework governing military applications of AI. Beyond that, much remains unclear. There was no shortage of ideas for where these discussions should go next, it now falls to the UN General Assembly to decide during its 81st session. There is an important window of opportunity to shape the direction we are heading in and to bring additional perspectives and expertise into any official process moving forward, including from relevant human rights actors across the UN system as well as non-governmental organisations and experts focusing on digital human rights. By taking human rights law as a starting point, incorporating the lessons and gains from broader AI governance frameworks, and fully reflecting the role and responsibilities of technology companies, future discussions would not be starting from zero. Rather, they would build on an existing foundation that can be adapted to provide the necessary protections where risks are highest: in the military context and in situations of armed conflict.
Leave a Reply