Ahead of the conference’s second edition, a series of three reports – the first of the AISE Markers series – takes stock of the key takeaways from the previous year and provides an initial basis and scaffolding for UNIDIR’s Global Conference on AI, Security and Ethics 2026. By acting as a bridge between editions, the AISE Markers series ensures that each conference is built on solid ground and constitutes a natural evolution of the discussions held in the previous conference.

This third report provides a structured account of where use cases – both governance and technological – stood in 2025. It offers an overview of discussions surrounding the human element in practice, surveys regional perspectives, zooms in on AI for peace operations and humanitarian applications, and then offers cross-cutting observations that may serve as a baseline for the 2026 edition. Accompanying reports will cover the state of governance and technology.

Citation: Yasmin Afina and Jan Hendrik Mannsperger, “AISE Markers Series – Benchmark III: Use Cases, Insights from the Global Conference on AI, Security and Ethics 2025”, Geneva: UNIDIR, 2026.

The post Benchmark III: Use Cases – Insights from the Global Conference on AI, Security and Ethics 2025 first appeared on UNIDIR.

Ahead of the conference’s second edition, a series of three reports – the first of the AISE Markers series – takes stock of the key takeaways from the previous year and provides an initial basis and scaffolding for UNIDIR’s Global Conference on AI, Security and Ethics 2026. By acting as a bridge between editions, the AISE Markers series ensures that each conference is built on solid ground and constitutes a natural evolution of the discussions held in the previous conference.

This second report provides a structured account of where the conversation on technology stood in 2025. It offers an overview of shared understandings – particularly in the areas of dual-use transformation, life cycle management, data and assurances – and identifies areas that may subsequently serve as a baseline for the 2026 edition. Accompanying reports will cover the state of governance and use cases.

Citation: Yasmin Afina and Jan Hendrik Mannsperger, “AISE Markers Series – Benchmark II: Technology, Insights from the Global Conference on AI, Security and Ethics 2025”, Geneva: UNIDIR, 2026.

The post Benchmark II: Technology – Insights from the Global Conference on AI, Security and Ethics 2025 first appeared on UNIDIR.

Ahead of the conference’s second edition, a series of three reports – the first of the AISE Markers series – takes stock of the key takeaways from the previous year and provides an initial basis and scaffolding for UNIDIR’s Global Conference on AI, Security and Ethics 2026. By acting as a bridge between editions, the AISE Markers series ensures that each conference is built on solid ground and constitutes a natural evolution of the discussions held in the previous conference.

This first report provides a structured account of where the conversation on governance stood in 2025. It offers an overview of shared understandings and possible areas of tension, and identifies issues that remain unresolved and may subsequently serve as a baseline for the 2026 edition. Accompanying reports will cover the state of technology and use cases.

Citation: Yasmin Afina and Jan Hendrik Mannsperger, “AISE Markers Series – Benchmark I: Governance, Insights from the Global Conference on AI, Security and Ethics 2025”, Geneva: UNIDIR, 2026.

The post Benchmark I: Governance – Insights from the Global Conference on AI, Security and Ethics 2025 first appeared on UNIDIR.

This compendium offers a comprehensive, neutral and technical resource for States, organized around realistic illustrations of how international humanitarian law (IHL) principles and obligations may be considered in practice. The scenarios explore situations of both international and non-international armed conflict, organized into three domains of warfare: air, land and naval domain. Each scenario is accompanied by questions regarding IHL compliance relevant to the (potential) use of LAWS in that context.

The scenario compendium provides a framework to consider operational and legal questions on the use of LAWS and it can be leveraged for two main purposes:

• At the national level, the compendium may support efforts to carry structured legal analysis, which can contribute to the development and refinement of IHL interpretation, as well as the stress-testing of existing, or the development of, new policies, tactics, rules of engagements, etc.
• At the international level, the compendium can be leveraged for capacity building and confidence building. First, the scenarios provide ready-made constructs to explore legal and operational questions, especially useful for States with limited modelling capabilities. Second, in international discussions, the compendium can serve as a common analytical reference to compare and discuss interpretations of IHL and identify areas of convergence and divergence.

Citation: Giacomo Persi Paoli, Ioana Puscas and Yasmin Afina, Scenario Compendium for the Legal, Operational and Technical Analysis of Lethal Autonomous Weapon Systems (Geneva: UNIDIR, 2026).

The post Scenario Compendium for the Legal, Operational and Technical Analysis of Lethal Autonomous Weapon Systems first appeared on UNIDIR.

Facilitated by UNIDIR, the consultations sought to build on the 2024 REAIM Regional Consultations and the 2023 and 2024 Summits, in addition to capturing evolution in national views and policies on responsible AI in the military domain, regional priorities and multi-stakeholder engagement over the year.

This report seeks to capture the main takeaways from the five regional consultations, summarizing participants’ views and a selection of UNIDIR’s observations. Specifically, these observations are centred around a number of common threads that ran through all of the regional consultations (even though minor adjustments were made for each regional event to factor in its respective local context and realities):

• National policies and practices
• Looking back (post-REAIM 2023 and 2024 reflections)
• Looking ahead (reflections for the 2026 REAIM Summit)

In addition, this report provides a comprehensive overview of some of the takeaways from the discussions held with the multi-stakeholder community. One key objective of the consultations, in acknowledgment of the importance of multi-stakeholder engagement, is to take stock of the views of regional representatives from industry, civil society, academia and research institutes, as well as regional and international organizations.

This report also looks into the operationalization of responsible AI principles across the life cycle of AI-enabled military capabilities through the lenses of assurances, incident response, crisis management and risk reduction.

The report then lays out States’ reflections on the REAIM journey three years on from the inaugural summit. It concludes by identifying substantive areas of priority that States wish to see further pursued, both within REAIM and beyond, before presenting a series of concrete recommendations for the road ahead.

Citation: Yasmin Afina, The Global Prism of Military AI Governance: Reflections from the 2025 Regional Consultations on Responsible AI in the Military Domain (Geneva: UNIDIR, 2026).

The post The Global Prism of Military AI Governance: Reflections from the 2025 Regional Consultations on Responsible AI in the Military Domain first appeared on UNIDIR.

This year, with the agenda turning 25, the factsheet provides ideas around how AI represents an opportunity and an obstacle for its realization. It also presents an analysis of the current state of the integration of emerging technologies like AI within WPS national and regional action plans.

The factsheet concludes by proposing further areas of action to better include AI in the WPS Agenda, making it fit for purpose amidst changing realities around peace and conflict.

Citation: Shimona Mohan (2025) “Factsheet: Artificial Intelligence and the Women, Peace and Security Agenda”, UNIDIR, Geneva.

The post Factsheet: Artificial Intelligence and the Women, Peace and Security Agenda first appeared on UNIDIR.

This has been prompted by initiatives outside the United Nations, such as the Responsible AI in the Military Domain (REAIM) summits and the Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy. These processes were fundamental in increasing awareness and served as incubators for policy action on the international stage. Against this backdrop and for many years, UNIDIR has contributed significantly to initiating and shaping national, regional and international discussions through its research, its capacity-building and its convening power.

The push for responsible AI in the military domain has opened new channels for dialogue among states. The shared recognition of AI’s disruptive potential, both positive and negative, has led to international discussions specifically about ensuring its safe and controlled development, deployment, and use.

The international community now has an opportunity to shape the future of international peace and security in the era of AI, putting principles of responsible AI at the core. Such engagement can build trust and mutual understanding, future-proofing the international peace and security architecture.

To further advance multilateral discussions on this new and rapidly evolving issue, it is crucial to clarify what “the military domain” means and entails; to survey key applications of AI in military settings in order to understand the associated opportunities; and to analyse the challenges and consider recommendations for policy development at all levels. This report addresses each of these aspects in turn, drawing on UNIDIR’s research and analysis on these topics over the years. It then proposes a 10-step road map towards effective national and international governance of AI in the military domain.

Citation: UNIDIR’s Security and Technology Programme. “Artificial Intelligence in the Military Domain and Its Implications for International Peace and Security: An Evidence-Based Road Map for Future Policy Action”. Geneva, Switzerland: UNIDIR, 2025.

The post Artificial Intelligence in the Military Domain and Its Implications for International Peace and Security: An Evidence-Based Road Map for Future Policy Action first appeared on UNIDIR.

Against this backdrop, UNIDIR conducted a project on working “Towards a Common Understanding of the Application of IHL to Emerging Technologies in the Area of LAWS”. Building on the momentum on this topic, UNIDIR’s primary objective was to take stock of the existing state of affairs and to capture existing views, positions and approaches – across sectors and across regions – to the application of IHL to LAWS. To this end, the Institute has drafted a separate background paper that summarizes publicly available views expressed by states, scholars and other experts participating in multilateral discussions on the applicability and interpretation of IHL with respect to the development, deployment and use of LAWS. To complement this research, UNIDIR conducted a series of bilateral and regional consultations between November 2024 and March 2025. In partnership with regional partners, regional consultations were held in The Hague, Brasília, Pretoria and Singapore. The consultations were designed to provide a platform for open discussions, knowledge and information sharing and the deepening of regional understandings on the intersection between IHL and LAWS. Held in-person under the Chatham House Rule, participants included government-affiliated experts in law, policy and defence from various ministries, national agencies and authorities, as well as a select number of scholars specializing in IHL and policy.

Citation: Yasmin Afina, “Regional Perspectives on the Application of International Humanitarian Law to Lethal Autonomous Weapons Systems”, UNIDIR, Geneva, 2025.

The post Regional Perspectives on the Application of International Humanitarian Law to Lethal Autonomous Weapon Systems first appeared on UNIDIR.

This briefing note will contribute to a report submitted to the 18th session of General Assembly and aims to support States in the formulation of their national views on this topic. It seeks to ensure that the resulting report is as comprehensive, diverse and geographically representative as possible. The brief includes some contextual information on the topic of AI in the military domain, a set of considerations for States to refer to, and a list of suggested readings that draws on UNIDIR’s own research and selected external publications.

Citation: Giacomo Persi Paoli and Yasmin Afina, “AI in the Military Domain: A briefing note for States”, UNIDIR, Geneva, 2025.

The post AI in the Military Domain: A briefing note for States first appeared on UNIDIR.

To support these ongoing discussions, UNIDIR implemented a series of activities as part of the project “Towards a Common Understanding of the Application of IHL to Emerging Technologies in the Area of LAWS”.  This background paper summarizes publicly available views expressed by States, scholars and other experts participating in multilateral discussions on the applicability and interpretation of IHL with respect to the development and use of LAWS.

The background paper finds that, while all contributions to the discussion stem from the common starting point that IHL applies to the development and use of LAWS, divergences of both form and content persist in publicly available views. Despite the breadth of the discussions, a coherent comparison of views remains difficult to achieve and some IHL rules that govern the development and use of LAWS remain underexamined. Publicized views on measures that States can, do or should take with respect to the development and use of LAWS to avoid or minimize the effects of LAWS on civilian populations, civilians and civilian objects often do not specify whether such measures derive from an IHL principle or rule. The background paper underscores the considerations that arise in ensuring that LAWS are developed and used only in accordance with IHL and the challenges in achieving a level of certainty about the interpretation and application of IHL to these technologies.

The post The Interpretation and Application of International Humanitarian Law in Relation to Lethal Autonomous Weapon Systems first appeared on UNIDIR.

• First, the development of AI technologies has – at least in part – depended on the availability of large datasets to train AI models.
• Second, data is a resource whose availability, collection, cleaning, use and sharing is affected by factors such as collection costs, lack of real-world data in certain domains, as well as regulatory, legal and ethical constraints.
• Third, data quality, representativeness, and diversity are directly linked to an AI model’s performance, level of bias, accuracy and reliability.

Synthetic data – data that is artificially generated in the digital world with properties that are often derived from an original set of data – has been proposed as a solution to address some of these data-related issues, especially for AI model training. However, synthetic data is no panacea, and has been shown to potentially exacerbate many of the issues it seeks to curtail, sparking governance discussions.

To explore the governance challenges of synthetic data in the context of international security, UNIDIR’s Security and Technology Programme held an event titled Technology and Security Seminar on Synthetic Data: Exploring Governance Implications.

This report provides a summary of the key themes and takeaways from discussions at the event. The report is divided into two parts, reflecting the structure of the event. The first part provides a short overview of the technology and its uses in the military domain. The second part presents the various views, issues and potential challenges to governance presented by synthetic data in the context of international security.

Citation: Federico Mantellassi, “Governance Implications of Synthetic Data in the Context of International Security”, UNIDIR, Geneva, 2024.

The post Governance Implications of Synthetic Data in the Context of International Security first appeared on UNIDIR.

This publication unpacks the AI-ICT security nexus and outlines, through a very easy-to-read, infographic the main current applications of AI for offensive and defensive purposes. To explain AI applications to the ICT environment, this study introduces the UNIDIR’s Intrusion Phases model, which is a framework that identifies three areas where AI can be used: outside the network perimeter, on the network perimeter, and inside the network perimeter.

Citation: Giacomo Persi Paoli, Samuele Dominioni. “Exploring the AI-ICT Security Nexus”. Geneva, Switzerland: UNIDIR, 2024.

The post Exploring the AI-ICT Security Nexus first appeared on UNIDIR.

Yet, most of these policies exclude or barely touch upon security and defence applications. Only a handful of national strategy documents have a section dedicated to this realm; and even fewer are specifically dedicated to it. This scarcity is at odds with the United Nations Secretary-General’s recommendation for Member States to “urgently develop national strategies on responsible design, development and use of artificial intelligence”, as outlined in his New Agenda for Peace.

Against this backdrop, UNIDIR has launched a programme of work to establish guidelines for the development, adoption, implementation and review of national strategies on AI in security and defence. The purpose of the guidelines, both of procedural and substantive nature, is to capture, anticipate and dissect the key issues, considerations and needs that each state must address as it develops or seeks to develop, adopt, implement and review its national strategy on AI in security and defence. In recognition of the host of incentives stemming from the establishment of such strategies, it is hoped that these guidelines will serve as a useful tool for states and non-state stakeholders alike as they seek to address issues related to the responsible development, deployment and use of AI in security and defence.

The present draft guidelines have been released to provide states and all relevant stakeholders involved in the development, adoption, implementation and review of national strategies on AI in security and defence with an opportunity to review and provide feedback to UNIDIR. The Institute aims to adopt a holistic and inclusive method to the establishment of the guidelines; it thus seeks to capture all the varying perspectives, viewpoints and approaches to this issue. We welcome feedback from stakeholders across sectors and domains.

The post Draft Guidelines for the Development of a National Strategy on AI in Security and Defence first appeared on UNIDIR.

The inaugural edition of RAISE convened participants primarily hailing from the industry and the research community, with select government representatives in light of the upcoming second edition of the Responsible AI in the Military Domain (REAIM) Summit (Seoul, 9-10 September 2024). Its objectives were two-fold:

1. Review the current state of applications of AI in security and defence contexts, across sectors and geographies but with a particular focus on the military domain; and
2. Identify key priority areas in which to develop specific guidance and policy recommendations on identified issues.

This inaugural edition of RAISE focused specifically on the military domain due to the momentum in this particular area. As a result of the convening and UNIDIR’s facilitation of the discussions, participants had identified, by its end, six key priority areas for RAISE to advance in the governance of AI in the military domain: building a knowledge base; trust building; the human element in AI uses; data practices; life cycle management; and destabilization.

This Policy Brief aims to lay the foundation for future work that will bring to life recommendations revolving around the six priority themes that the meeting identified, and which it agreed would serve as a basis for cooperation and collective action that transcends geopolitical rivalry, cross-sectoral divides and competition.

The post Governance of Artificial Intelligence in the Military Domain: A Multi-Stakeholder Perspective on Priority Areas first appeared on UNIDIR.

This report captures UNIDIR’s main reflections on the key takeaways stemming from the five regional consultations. These consultations did indeed enable the dissection of local contexts, realities and approaches with regard to the responsible development, deployment and use of AI in the military and wider security domains – including the identification of areas of nuanced convergence and divergence at the regional level.

Specifically, this report first discusses the reflections shared by States on the unique characteristics of AI technologies and the opportunities that they provide in the military domain. In addition, States also discussed and exchanged views on the risks, challenges and implications stemming from the development, deployment and use of AI in the military and wider security domains. The report then covers six points of convergence that have emerged from the consultations, along with five main points of divergence observed across and within regions.

The post The Global Kaleidoscope of Military AI Governance first appeared on UNIDIR.

It also addresses how biases in LAWS have been discussed at relevant disarmament forums like the Group of Governmental Experts (GGE) meetings under the Convention on Certain Convention Weapons (CCW).

The factsheet further recommends areas of action for a host of stakeholders to ensure that gender biases are mitigated in military applications of AI.

Citation: Gender and Disarmament & Security and Technology Programmes (2024) “Factsheet: Gender and Lethal Autonomous Weapons Systems”, UNIDIR, Geneva, Switzerland.

The post Gender and Lethal Autonomous Weapons Systems first appeared on UNIDIR.

The project consisted of two distinct phases:

1. The first developed a comprehensive taxonomy of AI risks in the context of international security.
2. The second provided a theoretical and historical overview of CBMs, probing for viable ways to move forward and seeking input from States.

This report concludes the second phase of the project and presents a framework for conceptual and practical considerations for CBMs for AI, drawing on lessons learned from other domains and from perspectives shared by a diverse group of States.

This study and the consultation UNIDIR convened with national representatives, which included a workshop and surveys, brought to light some key areas of agreement and shared concerns. As conversations about future CBMs begin to take shape, this publication provides a realistic assessment of current priorities and invites reflection on next steps.

The research output from the first phase of the project is available in the form of our earlier publication AI and International Security: Understanding the Risks and Paving the Path for Confidence-Building Measures.

The initial framing paper which launched the project is also available as Confidence-Building Measures for Artificial Intelligence: A Framing Paper.

Citation: Ioana Puscas, Confidence-Building Measures for Artificial Intelligence. A Multilateral Perspective, UNIDIR, Geneva, 2024.

The post Confidence-Building Measures for Artificial Intelligence: A Multilateral Perspective first appeared on UNIDIR.

This UNIDIR Primer provides an overview of the main opportunities and limitations of synthetic data in the training of AI systems. While synthetic data can be a proxy for real-world data and help shorten training cycles, among other benefits, there are also significant risks and challenges associated with its use.

The Primer explores existing data challenges, both technical and organizational, introduces key technical characteristics and methods of generating synthetic data, and analyzes implications of using synthetic data in the context of international security, including for autonomous systems and in the cyber realm.

Sponsor Organizations: The European Union; the governments of Czech Republic, Germany, Italy, Netherlands, Switzerland, and Microsoft.

Citation: Harry Deng (2023). “Exploring Synthetic Data for Artificial Intelligence and Autonomous Systems: A Primer”, UNIDIR, Geneva, Switzerland.

The post Exploring Synthetic Data for Artificial Intelligence and Autonomous Systems: A Primer first appeared on UNIDIR.

Advances in artificial intelligence (AI) in recent years, combined with the technology’s scalability and convergence with other domains, have prompted numerous concerns about the risks of AI to global security, including risks of misuse and escalation. However, policy discussions still lack a comprehensive analysis of the technology’s risks and how categories of risks are interrelated.

This report provides an overview of the main categories of risks of AI in the context of international peace and security, across domains of use and applications.

This research concludes phase one of the UNIDIR project on CBMs for AI. It provides a basis for multi-stakeholder engagements to understand the risks and to advance discussions about CBMs, which can help promote a more transparent, safe and responsible environment for the development and use of AI.

Citation: Ioana Puscas (2023) “AI and International Security: Understanding the Risks and Paving the Path for Confidence-Building Measures”, UNIDIR, Geneva, Switzerland

The post AI and International Security: Understanding the Risks and Paving the Path for Confidence-Building Measures first appeared on UNIDIR.

In the midst of discussions and debates around the opportunities and risks of AI for military purposes, as well as the governance and responsible use of these technologies, the United Nations Institute for Disarmament Research (UNIDIR)’s new report aims to increase understanding of the role of AI in the execution of military tasks beyond applications relating to the use of force and the narrow tasks of target selection and target engagement within the targeting process.

The report provides an overview of current and near-future AI capabilities relevant to aiding with 18 military tasks. The paper also presents a discussion on the strengths and limitations regarding the application of AI to these military tasks.

Citation: Sarah Grand-Clément, “Artificial Intelligence Beyond Weapons: Application and Impact of AI in the Military Domain”, UNIDIR, Geneva, 2023.

The post Artificial Intelligence Beyond Weapons: Application and Impact of AI in the Military Domain first appeared on UNIDIR.

Part I of the report seeks to provide a foundational understanding of the concept of AI and its state of play.

Part II examines the disruptive impact of AI on international peace and security. In particular, it discusses the risks and benefits of uses of AI in military operations and across domains of warfare as well as the opportunities and challenges of harnessing AI technologies for conflict prevention and peacebuilding.

Part III of the report examines the path to Responsible AI. It unpacks the RAI governance approach and discusses how it is and can be operationalized. It also reflects on the value of building an RAI culture.

HIGHLIGHTS & RECORDINGS:

• Read a brief summary of the Conference Highlights
• Watch all of the conference sessions again (via the UNIDIR YouTube channel or below)

Citation: Wenting He and Alisha Anand (2023) “The 2022 Innovations Dialogue: AI Disruption, Peace and Security”, UNIDIR, Geneva, Switzerland.

The post The 2022 Innovations Dialogue: AI Disruption, Peace and Security (Conference Report) first appeared on UNIDIR.

UNIDIR’s project Towards Responsible AI in Defence seeks to, first, build a common understanding of the key facets of responsible research, design, development, deployment, and use of AI systems. It will then examine the operationalisation of Responsible AI in the defence sector, including identifying and facilitating the exchange of good practices. The project has three main aims. First, it aims to encourage states to adopt and operationalise tools that can enable responsible behaviour in the development and use of AI systems. It also seeks to help increase transparency and foster trust among states and other key AI actors. Finally, the project aims to build a shared understanding of the key elements of Responsible AI and how they may be operationalised, which may inform the development of internationally accepted governance frameworks.

This research brief provides an overview of the aims of the project. It also outlines the research methodology for and preliminary findings of the project’s first phase: the development of a common taxonomy of principles and a comparative analysis of AI principles adopted by states.

Citation: Alisha Anand and Harry Deng (2023) “Towards Responsible AI in Defence: A Mapping and Comparative Analysis of AI Principles Adopted by States”, UNIDIR, Geneva, Switzerland.

The post Towards Responsible AI in Defence: A Mapping and Comparative Analysis of AI Principles Adopted by States first appeared on UNIDIR.

Confidence-building measures (CBMs) for AI can provide flexible options for the future development and deployment of AI-enabled systems. This framing paper introduces a new UNIDIR project, which aims at developing a possible roadmap for the future elaboration of CBMs for AI.

The first phase of the project consists of a risk-mapping analysis, unpacking risks of the technology and assessing how they may translate into risks for international peace and security. The second phase of the project will consider pathways for the elaboration of CBMs in a series of multi-stakeholder engagements.

Citation: Ioana Puscas (2022) “Confidence-Building Measures for Artificial Intelligence: A Framing Paper”, UNIDIR, Geneva, Switzerland.

The post Confidence-Building Measures for Artificial Intelligence: A Framing Paper first appeared on UNIDIR.

This study provides a clear, comprehensive introduction to these questions, and proposes a range of avenues for action by which they may be addressed.

Citation: Arthur Holland Michel (2020) “The Black Box, Unlocked”, UNIDIR, Geneva. doi: 10.37559/SecTec/20/AI1

Teaser: Predictability and Understandability in Military AI

The post The Black Box, Unlocked first appeared on UNIDIR.