AT three o’clock on the morning of Saturday, Sept 5, 2026, Room III of the Palais des Nations in Geneva — the only room in the building with any sign of life — erupted in applause. The Group of Governmental Experts on Lethal Autonomous Weapons Systems (LAWS), under the able leadership of its Dutch chair, had adopted — after three years of negotiations — a report containing elements of an instrument. On the final day, the delegations of the 128 high contracting parties to the Convention on Certain Conventional Weapons (CCW) sat for 18 hours without a break.
The report contains, for the first time, an agreed characterisation of LAWS. The subject has been under discussion in the CCW since 2014, and only now have states parties been able to say which systems fall under the label. The report also reaffirms that control and human judgement are needed to ensure compliance with international humanitarian law (IHL), that human beings must determine the measures necessary for that compliance. It sets out a non-exhaustive list of what those measures might be.
Part of the applause that morning was relief, and part of the relief was that the text had survived at all. Over three years, a small number of militarily significant states, the very states developing and fielding these technologies, worked steadily to thin it out. ‘Human control’ became ‘judgement and control’, and then control that could be exercised indirectly. Predictability, reliability, and explainability, once framed as properties a system must possess, were dropped from the text altogether. A proposed requirement that a system must not alter its target parameters or its mission without human authorisation did not make it to the final agreement either.
No new prohibitive norm survived. An honest reading of the text is that it gives a green light to the development and use of LAWS, provided certain measures are in place. Those measures are real, and they are better than nothing. But they focus on how the existing law applies to this category of systems. They fall well short of the joint appeal issued in October 2023 by the UN secretary-general and the president of the International Committee of the Red Cross, who called on states to conclude by 2026 a legally binding instrument prohibiting autonomous weapon systems that operate without human control or that target human beings directly, and regulating all others. A handful of militarily significant states continue to resist any negotiation of a new legal instrument and are unlikely to accept one.
Technology on the battlefield has always outpaced the human ability to govern it.
Back in 2014, when the discussions first began, autonomous weapons were largely a matter of science fiction. They are not any more. Algorithmic decision-support systems have been used to generate targets at a speed no human can match. In various conflicts, loitering munitions and drones with autonomous terminal guidance have been used, precisely because they provide an advantage, are cheap, and keep working when the electronic warfare environment cuts the link to the operator. Such systems are also no longer prospective in our own region; they have been inducted into arsenals and recently used. These developments in the battlefield did not wait for the Geneva diplomatic community to settle first on a characterisation of LAWS or to spell out how the existing IHL applies to such systems.
In 2013, Pakistan was among the first states to call for a legally binding prohibition on systems that decide on the use of force without human control and that do not allow a human user to understand, predict, and explain their decisions. As the technology evolved and was deployed on the battlefield, Pakistan also adapted its position to argue for a two-tier approach, with certain prohibitions in cases where IHL compliance could not be ensured, and restricting the use in all other cases. In its proposals submitted in 2023 and 2024, Pakistan laid out elements of a new legal instrument under this two-tier approach. It argued that any prohibition must cover development and not only use, since a weapon that is unlawful to use should not be lawful to build. Our position in support of a legal instrument found resonance and came to be shared by a majority of states, growing into a group of more than 100. An argument advanced by a handful of delegations 12 years ago has become the majority voice today.
Technology on the battlefield has always outpaced the human ability to govern it. The record of arms control bears this out. Prohibitions have been agreed only where the weapon had not yet been fielded, as with blinding laser weapons, or where it offered no decisive advantage to the states that possessed it, as with biological weapons. It is difficult to find instances in which major military powers negotiated away the advantage they held on the battlefield. Restraint has been agreed for weapons that were marginal, obsolete, or not yet real.
Autonomy in weapons is none of those things. For major powers, it has become the organising principle of the next generation of military doctrine and the largest defence-industrial investment of recent years. For other states, it has also become an instrument to offset asymmetries in different domains of warfare. It is unlikely that such states will accept meaningful constraints on these capabilities.
While Pakistan will continue to argue for restraints, no state can be expected to rest its security on rules that do not yet exist, or on rules by which its adversaries have chosen not to be bound. Pakistan will keep a clear view of that reality and will need to take steps accordingly.
The writer is a Pakistani diplomat, serving as counsellor in the Permanent Mission of Pakistan to the United Nations in Geneva and participating in the negotiations on the Lethal Autonomous Weapons Systems.
Published in Dawn, October 24th, 2026
