UNIDIR played a key role at the Third Summit on Responsible Artificial Intelligence in the Military Domain (REAIM), held from 4-5 February in A Coruña, Spain. The summit brought together State representatives, international organizations, industry, academia and civil society to advance responsible approaches to military AI.
Through a series of dedicated sessions, UNIDIR shepherded discussions reflecting its ongoing work on AI and international peace and security. The Institute led on exchanges to advance the operationalization of responsible AI principles – a priority identified in the summit’s outcome document.
A longstanding commitment to responsible AI
UNIDIR’s participation at REAIM 2026 builds on sustained contributions to the initiative. The Institute has played an advisory role to successive host States (the Netherlands, the Republic of Korea, and Spain) – well before 2023, helping inform REAIM’s evolution.
This leadership has extended beyond intergovernmental settings, through the design and delivery of the 2025 REAIM Regional Consultations. It also facilitated structured multi-stakeholder engagement, as well as expert contributions to the Global Commission on Responsible AI in the Military Domain.
Rooted in independence, technical depth, and constant engagement, UNIDIR continues to translate evidence into policy-relevant pathways and solutions – particularly as States face governance challenges around the military implications of AI.
Regional dialogue and partnerships at the forefront
In the margins of REAIM 2026, UNIDIR Director Robin Geiss met with Kenya’s Cabinet Secretary for Defense, Hon. Soipan Tuya, reaffirming and deepening cooperation between the Institute and Kenya on responsible AI in the military domain. The exchange reflected a shared commitment to advancing responsible approaches to military AI and underscored the importance of trust-based, inclusive partnerships at both regional and global levels.
Kenya’s leadership was also evident as the host of the 2024 and 2025 REAIM Regional Consultations for Africa, which highlighted the value of inclusive regional dialogue and locally grounded perspectives in shaping global governance debates on emerging technologies and international security.
In this vein, UNIDIR launched its summary report capturing the perspectives shared at the 2025 REAIM Regional Consultations. The Global Prism of Military AI Governance presents existing national policies and best practices for the governance of AI in the military domain, as well as views from discussions held with multi-stakeholder communities across regions.
The report looks into the operationalization of responsible AI principles across the life cycle of AI-enabled military capabilities through lenses of procurement, incident response, crisis management and risk reduction. It offers quantitative data on assurance prioritization for the purchase of different military AI capabilities. The publication also lays out States’ reflections on the REAIM journey – three years from the inaugural summit. It concludes by identifying substantive priority areas that States wish to see further pursued, both within REAIM and beyond, before presenting a series of concrete recommendations for the road ahead.
UNIDIR’s Security and Technology Programme also met with representatives from the Center for International Security and Strategy (CISS) of Tsinghua University. The discussions highlighted the value of partnerships with leading academic institutions and built on ongoing collaboration through UNIDIR’s Roundtable for AI, Security and Ethics (RAISE). The meeting also constituted an opportunity to exchange on good practices of confidence-building measures on AI in the military domain, emphasizing the work of CISS Tsinghua University’s US-China Track-2 Dialogue on AI and International Security with the Brookings Institution, and offered pathways for further cooperation with UNIDIR.
Engaging industry on military AI
Industry engagement was another central focus for UNIDIR at REAIM 2026, reflecting the increasingly decisive role of private sector actors in shaping AI-enabled military capabilities. From system design and data practices to testing, deployment and post-deployment support, industry decisions have a direct impact on reliability, predictability, accountability and the risk of harm in real-world contexts. As governance efforts in the military AI domain advance, effective collaboration with industry has become critical to translating shared principles into practice.
At the summit, UNIDIR – in partnership with the Office of the High Commissioner for Human Rights (OHCHR) – launched the Framework of Responsible Industry Behaviour for AI in the Military Domain. This sets out a practical and actionable set of voluntary guidelines aimed at providing a common baseline for responsible industry conduct. The framework offers governments and other stakeholders greater clarity on responsible AI procurement and public-private partnerships.
Ahead of the launch, UNIDIR and the OHCHR convened a closed-door roundtable with industry representatives attending the summit. They were joined by framework collaborators from Microsoft, the Japan Defense Technology Foundation and Hitachi America, Ltd. The exchanges, held under the Chatham House rule, provided space for the private sector to share perspectives on substantive and project-oriented themes. They explored risk perceptions and practical strategies to address them, misuse pathways and assurance gaps, priority areas for governance, and the value of the framework at this point in time.
What’s next for military AI governance
During his remarks at the high-level plenary session, UNIDIR Director echoed what many participants at the summit have noted: REAIM is at an inflection point. The summit has been instrumental in catalyzing global momentum on responsible military AI. The 2024 and 2025 REAIM Regional Consultations, in particular, have translated international debate into regional and national relevance.
Yet, as UN processes are now growing, future efforts for the governance of AI in the military domain must be approached with coherence in mind. Coherence will be key for the effectiveness and optimization of available resources, and to ensure that ultimately, the development and deployment of AI will foster international peace and security – not undermine it.
With proper governance in place, military AI could help reduce harm to civilians by improving situational awareness, highlighting uncertainty, and supporting more careful use of force.
Robin Geiss, UNIDIR Director
UNIDIR’s participation at REAIM 2026 marks the beginning of much work to build on the discussions held in A Coruña, with a host of opportunities for States, the private sector and the wider multi-stakeholder community to contribute meaningfully to the Institute’s work.
UNIDIR is actively on the lookout for collaborators to support the development of the Framework of Responsible Industry Behaviour for AI in the Military Domain. Collaboration does not imply endorsement or entail binding commitments, but industry participation in the development process will be critical to the framework’s ultimate utility and effectiveness.
Looking ahead, UNIDIR will organize the second edition of its Global Conference on AI, Security and Ethics on 18-19 June 2026. This will be held consecutively to the informal exchange in Geneva on AI in the military domain and its implications for international peace and security, organized pursuant to the UN First Committee Resolution 80/58. UNIDIR welcomes the submission of proposals to hold thematic deep-dives, lightning talks and poster presentations, the details of which will be released in late February.
