Key takeaways
- Lessons from cyberspace show that sustained, inclusive and incremental norm-building is possible even when States disagree on major political questions.
- Future military AI norms could combine positive commitments, such as protecting critical infrastructure, with negative commitments, such as restraints on particularly high-risk applications.
- Developing countries need greater capacity-building and technical cooperation to participate meaningfully in international discussions on military AI.
- AI incident reporting, points of contact and emergency response mechanisms could provide practical tools for transparency and confidence-building measures.
This publication examines what lessons the international community can draw from the development of voluntary norms of responsible State behaviour in information and communications technology (ICT) security for future discussions on AI in the military domain. It argues that norms in ICT security should not be transferred directly to military AI but can serve as useful signposts when adapted to AI’s distinctive characteristics, including its dependence on data and autonomous functions.
The publication identifies growing convergence among States views around the applicability of international law, the need to maintain control and human judgment over the use of force, accountability, and protection of civilians. At the same time, important differences remain over terminology, red lines, voluntary versus legally binding measures, and next steps in multilateral processes.
Drawing on ICT security norms, the publication highlights possible areas for future action, including cooperation, prevention of misuse, respect for international law, protection of critical infrastructure, incident response and private sector engagement. It concludes that the pace of AI development requires multilateral discussions to advance more rapidly, while further work is needed on industry engagement, confidence-building measures and capacity-building.