Diplomats gathered at the United Nations in Geneva this week to negotiate possible rules governing lethal autonomous weapons, systems often described as killer robots, as pressure mounts to act before the technology moves beyond the reach of international law.

The conference on Certain Conventional Weapons, known as the CCW, convenes only once every five years, and supporters of new restrictions say the session may be the last realistic opportunity to agree on limits before fully autonomous weapons arrive.

Military technology has advanced rapidly in recent years, and some experts caution that the world could be only months away from weapons capable of selecting targets and deciding to kill on their own. AI-assisted weapons have already been used to strike targets in Ukraine and Gaza, where Ukrainian civilians have reported being 'hunted' by drones while shopping or going to school, though those systems remain under remote human control.

Humanitarian organizations insist that the decision to take a life, and responsibility for that choice, must remain with people. Ahead of the meeting, the United Nations, the International Committee of the Red Cross and the Pope all called for legally binding rules, and U.N. Secretary-General Antonio Guterres warned that allowing fully autonomous weapons to kill would mean 'crossing a moral red line.'

The Red Cross is especially concerned that autonomous systems cannot fully respect the Geneva Conventions, which bar targeting soldiers who are wounded or surrendering in the chaos of a fast-changing battlefield. Georgia Haines, a legal adviser with the organization, said machines cannot read contextual signals the way humans can, warning that 'the risks are incredibly high that there will be a misidentification of a target.'

Human rights groups add that the technology behind AI-assisted weapons is becoming increasingly cheap and easy to use, raising the danger it could spread to terrorist groups or to law enforcement agencies in repressive regimes. In theory, they warn, a swarm of autonomous drones could be programmed to kill everyone wearing a red sweater, everyone carrying a poster or even everyone speaking a certain language.

At least 50 countries, including Ireland, Italy, Brazil, Mexico, France and Germany, are pressing for a strict timetable to agree on binding rules. But states already developing and using semi-autonomous weapons are pushing back: the United States, Russia and Israel have signaled they will oppose legally binding rules, with Washington proposing a five-year timetable to draft non-binding guidelines instead.

The United States has traditionally been a late adopter of international treaties. It never ratified the Ottawa Convention on landmines, for example, though over time it stopped deploying them and reduced its stockpiles. Doubts about international law have grown under President Trump, whose administration has imposed sweeping sanctions on the International Criminal Court, and over the weekend Secretary of State Marco Rubio used a keynote speech to the Federalist Society to warn against 'quoting laws to men with swords.'

Complicating matters, U.N. negotiations on limiting specific weapons are meant to proceed by consensus. That approach delivered a ban on blinding lasers in 1995, but other agreements have proved elusive; frustrated states eventually took landmine talks outside the U.N. process to create the Ottawa Convention.

A compromise may emerge this week in the form of a longer or even open-ended negotiating period to produce rules that individual states could adopt as binding or non-binding. Diplomats backing that approach see it as a way to ensure at least some progress, while campaigners view it as postponing action on an urgent and dangerous problem.