
Killer robots: The race to control the weapons that can decide who lives or dies: NPR
A Ukrainian FP-1 long-range attack drone is launched at an undisclosed location in Ukraine, August 16, 2026.
Efrem Lukatski/AP
hide caption
Efrem Lukatski/AP
GENEVA — Diplomats meeting at the United Nations in Geneva have agreed to begin discussing options to regulate deadly autonomous weapons, often called “killer robots.”
After a week of negotiations that stretched into the early hours of Saturday morning, a compromise emerged: Countries would have two years to propose regulations that could be legally binding, but could also be optional.
Tear-eyed negotiators insisted the deal was better than nothing and demonstrated the U.N. disarmament process could still work. But for many aid organizations and countries, the deal falls far short of what they hoped for.
Before negotiations began, the International Committee of the Red Cross, guardian of the Geneva Conventions, said there was an urgent need to develop legally binding rules within a year. This call was supported by the UN Secretary General and Pope Leo.
Their concerns stem from the type of war we are currently witnessing. In Ukraine, Sudan and Gaza, civilians are being killed by semi-autonomous drones. This year in Sudan, 80% of civilian deaths were caused by drones according to the UN

Humans remain in control for now, albeit very distantly, but aid organizations fear we are on the verge of seeing AI-assisted weapons make the decision to kill entirely on their own.
Algorithms making life or death decisions
“There is a fundamental ethical concern that essentially life and death decisions are determined by programming, by an algorithm,” said Georgia Hinds, legal adviser to the ICRC. “This is not something that can be solved with better technology. It is a moral red line that we cannot cross.”
No one went into negotiations thinking they would be easy; some of the world’s most powerful countries, including the United States, Russia and Israel, had already expressed reluctance to accept binding regulations.
From the first day, Russia delayed the work by opposing the meeting’s agenda. Next, the United States proposed a five-year window to discuss voluntary, non-binding guidelines for autonomous weapons.

“They just throw away the pieces,” said one frustrated delegate. As time passed and the weekend approached, diplomats raced to get some sort of outcome from negotiations that had been billed as the last chance to regulate weapons whose technology and decision-making capabilities we may not even fully understand.
Finally, on Saturday morning at 4:30 a.m., they emerged from the negotiating room to announce their compromise: no to an agreement within a year, maybe to legally binding rules, or maybe not.
A “nothing burger”
The ICRC immediately said it regretted the decision and renewed its call for a binding treaty. Other activists were less polite, calling the deal a “real disappointment” and a “nothing burger.” “I fear people will die because of this,” added another.
One of the main concerns is that, with two years to discuss possible rules that might only be optional, the technology will advance unchecked.
“We will likely see these systems proliferate,” said Anna Hehir, an expert in military AI governance at the Future of Life Institute. “They are unlike nuclear weapons in the sense that they are inaccessible. There are autonomous weapons that are easy to purchase or even create for all kinds of actors: individuals, non-state actors, criminal groups.”
There are also big questions about how our existing rules of war – the Geneva Conventions – can be respected if those doing the killing are not humans, but machines.
“These rules are the bare minimum that militaries and states have developed and agreed to follow,” said the ICRC’s Hinds. “The commander remains responsible for ensuring that the use of force is lawful. The commander remains responsible. You cannot hold a machine responsible for a violation of the law.”
But in the race to develop new weapons, no country wants to come second. Even though most UN member states agree that there should be some controls on autonomous weapons, the Geneva compromise means that, for the next two years at least, weapons capable of using artificial intelligence to make life-or-death decisions will not be regulated.
Meanwhile, weapons technology is developing rapidly, while diplomacy struggles to keep pace.
Gn world