UN Review Conference in Geneva to Decide Next Steps on AI-Powered Autonomous Weapons

According to the UN Office at Geneva, a Group of Governmental Experts that met from 2024 to 2026 adopted a final report setting out a "set of elements of an instrument" on these weapons. The report does not say what form that instrument should take. The Seventh Review Conference, which runs until November 20 under the presidency of Swiss Ambassador Julien Thöni, is expected to consider and endorse those elements and set the mandate and arrangements for further work. The UN Secretary-General and the International Committee of the Red Cross have jointly asked states to negotiate a legally binding instrument.

The outcome will decide whether weapons that use AI to pick and strike targets are covered by binding international rules, or whether they stay under national policies that currently differ. The convention works by consensus, so a single state can block a decision to open negotiations.

Where Governments Draw the Line

Jean-Marc Rickli, head of global and emerging risks at the Geneva Center for Security Policy, told AFP that such systems "can identify, select, and engage a target without the intervention of a human operator". He said that with the spread of AI-enhanced drones, they can now go after targets that were not chosen in advance and adapt to their surroundings without human input. The main question for regulators is how much human control must remain over the decision to use force.

National positions differ. The United States speaks of "appropriate levels of human judgment". France bans fully autonomous lethal systems but allows partially autonomous ones. Rear Admiral Vincent Sebastien, deputy director of the French defence ministry's Agency for Defense Artificial Intelligence, said a military commander's "appropriate control" varies with the system and the operational context.

In a joint statement in August, UN Secretary-General Antonio Guterres and ICRC president Mirjana Spoljaric warned: "We are now dangerously close to crossing a moral red line: the autonomous targeting of humans by machines."

Targeting Tools Outside the Talks

Some analysts say the Geneva agenda leaves out the fastest-growing military uses of AI. Rickli called its scope "very narrow" because it does not cover AI in analysis and target generation. He pointed to the US-Israeli war against Iran this year, in which he said US forces engaged as many as 1,000 targets a day. Laure de Roucy-Rochegonde of the French Institute of International Relations said many of the risks lie in targeting and decision-support systems that are already widely deployed. She asked what level of supervision is needed so that operators exercise real judgment rather than rubber-stamping a machine's recommendations.

Analysts say fully autonomous weapons are not yet in widespread use, though they have been deployed by Russia in Ukraine and in US attacks in Iran. Sebastien said autonomy "will very quickly become a reality on the battlefield". China's defence ministry described swarms of autonomous drones in January as the "new face of warfare". Gerald Mako, a research affiliate at Cambridge University, wrote in Small Wars Journal that the first state to normalise fully autonomous lethal strikes "will reset expectations for everyone else".

Industry has raised doubts of its own. Anthropic chief executive Dario Amodei warned this year that "frontier AI systems are simply not reliable enough to power fully autonomous weapons."

The conference's main decision is whether to give the convention a mandate to negotiate on the experts' draft elements. A consensus to do so would turn years of expert talks into formal treaty work. Without it, the issue could move to other forums.