From the drone battlefields of Ukraine to AI targeting in the Iran war, artificial intelligence is forcing humanity to renegotiate war itself. As machines acquire the capability to make decisions on the battlefield, we now face a question that we have never had to confront before: Should a machine be able to kill by itself, without any human control?
At the center of this debate are a new kind of weapons system, known as lethal autonomous weapons systems (LAWS), which are more commonly called ‘killer robots’.
As diplomats from across the world prepare to meet in Geneva next week for the Seventh Review Conference of the Convention on Certain Conventional Weapons (CCW), pressure is mounting on the United Nations to establish international rules governing weapons that can identify, select and engage targets without human intervention. The conference will take place from November 16 to 20 at the UN Office at Geneva, Switzerland.
The CCW is an international treaty and law that prohibits and restricts the use of certain conventional weapons considered excessively injurious or indiscriminate to combatants and civilians. Since 2017, the Group of Governmental Experts (GGE) on LAWS under the CCW have been examining one of the most consequential questions facing artificial intelligence, international law and national security: how should the world regulate weapons capable of autonomously killing human beings?
On September 4, 2026, the GGE adopted its final report, outlining elements for a possible international instrument regulating autonomous weapons. It marks an important diplomatic milestone, but it is not a binding treaty.
The next step and decision now rests with countries meeting in Geneva.
Can years of negotiations finally translate into binding international rules, before autonomous weapons become entrenched in modern warfare?
Defining the Weapon
Even before governments can regulate autonomous weapons, they must first agree on a definition.
The GGE describes a lethal autonomous weapon system as one capable of identifying, selecting and engaging a target in a way that results in loss of life, without human intervention.
Importantly, according to the GGE, pre-programming potential targets does not, by itself, take a weapons system outside this definition. However, where military personnel determine the specific target and timing of an attack, the system falls outside the agreed characterization.
The final report explicitly links autonomous targeting capabilities to loss of life. This narrows the scope of future regulation, making the precise definition particularly important in determining which technologies are covered as LAWS and which are not.
Keeping Humans in Control
Yet defining autonomous weapons is only the beginning. The more difficult question is how much control humans should retain over their use.
The GGE’s final report reaffirms that international humanitarian law applies fully to autonomous weapons.
The principles of distinction, proportionality and precaution in attack, remain essential safeguards for civilians during armed conflict.
But establishing that the law applies does not guarantee compliance when machines are selecting, engaging and killing targets.
The report makes clear that human control and judgement are necessary, whether exercised before or during an operation.
It recommends maintaining responsible chains of command, limiting the scale and geographical scope of operations, restricting target types and ensuring that autonomous weapons systems can be intercepted or deactivated where feasible.
The challenge is to make human control meaningful, rather than merely formal. Because a machine cannot be held responsible for taking a human life.
The GGE emphasises that states and individuals remain accountable under international law and that responsibility of killing people cannot be transferred to machines.
It also calls for adequate testing, personnel training and safeguards against bias in the AI models used in these weapons.
The report further highlights risks including civilian casualties, unintended engagements, loss of control and proliferation. It cautions against automation bias, the tendency of human operators to place excessive confidence in a machine’s recommendations.
Pressure for a UN Treaty
Globally, the push for a legally binding international treaty on LAWS is gaining momentum.
On August 25 this year, UN Secretary-General António Guterres and President of the International Committee of the Red Cross, Mirjana Spoljaric called for clear prohibitions and restrictions on autonomous weapons, warning against a future in which machines could target human beings without meaningful human control.
Their appeal reflects a growing concern among international institutions and humanitarian organisations that technological developments are advancing faster than the diplomatic and legal processes intended to regulate them.
And this is where the meeting in Geneva next week becomes critical.
The Seventh Review Conference will determine the future direction of the negotiations. Governments could agree to begin formal negotiations on a legally binding instrument, extend the GGE’s mandate for further discussions or fail to reach consensus on a clear next step.
It is evident that agreement will not come easily.
Countries still differ over the military advantages of autonomous weapons and the extent to which they are prepared to restrict their development and deployment.
Even a decision to begin negotiations would open the door to difficult discussions about what should be prohibited, which safeguards should be mandatory and how compliance should be ensured.
The GGE has nevertheless established important common ground: stating clearly that international humanitarian law applies to LAWS and tht human judgement matters.
It has also said that accountability cannot be delegated to machines.
The Cost of Waiting
Countries everywhere now face a choice.
They can establish the rules before autonomous weapons become an entrenched part of modern warfare, or attempt to impose restrictions after the technology has moved beyond control.
Ultimately, it is about deciding whether humanity is prepared to keep control over the most consequential decision in warfare: the decision to take a life.
Common sense would suggest that the answer is obvious. But in international politics, what is obvious is not always what is agreed.
The world will find out in Geneva.
