Overview
- Delegates from 128 states met in Geneva on Monday, November 16 to decide whether to launch formal negotiations on a treaty to regulate or ban AI systems that can identify, select and engage targets without human intervention.
- The United States has proposed a five-year process focused on non-binding guidelines and Russia has signalled opposition to binding limits, while a chair’s plan for a two-year mandate and a seven-state proposal seek a path to a legally binding protocol.
- UN leaders, the International Committee of the Red Cross, the pope and rights groups say fully autonomous targeting crosses a moral red line and warn that unpredictable AI errors threaten civilian safety and accountability.
- Under the Convention on Certain Conventional Weapons any single state can block agreement, which campaigners say helps powerful weapons makers stall progress and may produce watered-down or hybrid outcomes rather than clear prohibitions.
- If consensus fails negotiators will likely leave Geneva with voluntary measures or a diluted mandate, which experts warn could speed an arms race and leave civilians exposed as semi-autonomous and early autonomous systems gain battlefield use in conflicts such as Ukraine and the Middle East.