Artificial intelligence is becoming increasingly relevant to international peace and security. Rapid advances in AI capabilities and their broad accessibility create opportunities for development and innovation, but they also increase the risk that powerful tools may be acquired, adapted or misused by actors seeking to cause harm. Of particular concern is the potential use of AI by malicious non-State actors to support cyber operations, disinformation, autonomous or remotely enabled systems, and other activities with destabilizing consequences.

Existing non-proliferation regimes, expert networks and export-control frameworks offer a valuable foundation for addressing some of these risks. However, AI differs from many traditionally controlled technologies: it is largely software-based, develops rapidly, relies on globally distributed infrastructure and expertise, and often has legitimate civilian and commercial applications. These characteristics complicate conventional approaches to controlling access, transfer and end use.

Dedicated dialogue is therefore needed to identify where existing instruments remain useful, where gaps persist, and how future measures can balance security concerns with the benefits of peaceful and responsible AI development. Convened during the First Committee, this side event, organized by UNIDIR and UNITAR, will support a multidisciplinary exchange focused on realistic and implementable policy options.