WASHINGTON, September 20 — Jensen Huang, the chief executive of the chipmaker Nvidia, has said there is a “0% chance” that AI destroys the world by 2030. The claim, whatever you make of the arithmetic, has reignited a debate that has been grinding through defence ministries and UN committees for years: what happens when the machine picks the target itself.
Lethal autonomous weapons systems — LAWS, or AWS depending on which acronym your ministry prefers — are military drones or military robots that can independently search for and engage targets based on programmed constraints and descriptions. The technical reality is narrower than the phrase suggests, at least for now.
What is actually fielded
As of 2025, most military drones, including unmanned aerial vehicles and unmanned combat aerial vehicles, and most military robots, are not truly autonomous. Operators run them from remote terminals over radio links. Because no onboard human pilot needs supporting, the airframe sheds weight and size compared with a manned aircraft — the engineering dividend that made the whole category viable in the first place.
Where autonomy does creep in is at the edges. A UCAV, also called a combat drone, fighter drone or battlefield UAV, carries aircraft ordnance — air-to-surface missiles, anti-tank guided missiles, aerial bombs — on hardpoints or in a weapons bay, and performs tactical attacks known as drone strikes.
Recent work on Turkey’s and Australia’s UCAV programmes has demonstrated some multi-role combat characteristics, combining beyond-visual-range air-to-air combat with ground-attack capability. That is a meaningful step, and it is not the same thing as a machine deciding for itself whom to kill. UCAVs are used for intelligence, surveillance, target acquisition and reconnaissance, for attacking high-value targets, and for prolonged loitering before they need to return to base.
They are distinct from loitering munitions and one-way attack drones, which are single-use and built to ram a target and explode on impact, and from surveillance drones, which are unarmed and used only for aerial reconnaissance and intelligence gathering. In March 2026 the British government indicated it had aerial minesweeping drones and was considering deploying them in the Strait of Hormuz.
The definitional muddle
Here is the detail that matters. There is no definition of lethal autonomous weapon systems generally agreed upon among different countries, and the UN Secretary-General’s report to the General Assembly notes some of the definitional problems. Some observers argue the word “lethal” should not be used at all, on the grounds that lethality is an effect of the manner in which a weapon is used rather than an innate quality of the weapon.
The meaning of “autonomous” is disputed too, because existing systems commonly described as autonomous contain varying degrees of it. The US Department of Defense Policy on Autonomy in Weapon Systems, Directive 3000.09, defines an Autonomous Weapon System as one that, once activated, can select and engage targets without further intervention by a human operator.
That definition is more extensive than the UN Group of Governmental Experts’ version. The GGE calls a LAWS an integrated combination of one or more weapons and technological components that enable the system to identify and/or select, and engage a target, without intervention by a human user.
Directive 3000.09 matches the GGE on the core wording but adds that its definition includes, but is not limited to, operator-supervised autonomous weapon systems designed to let operators override operation of the weapon system, yet able to select and engage targets without further operator input after activation.
Heather Roff, writing for Case Western Reserve University School of Law, describes autonomous weapon systems as capable of learning and adapting their functioning in response to changing circumstances in the environment in which they are deployed, and capable of making firing decisions on their own. The British Ministry of Defence draws the line differently: an automated system can carry out complicated tasks but is incapable of complex decision-making, whereas an autonomous system is capable of deciding a course of action.
LAWS may engage in drone warfare in the air, on land, on water, underwater or in space. The risk that concerns the committees is straightforward — misidentification. Watch the GGE process and national procurement language for whether anyone converges on a shared definition, because without one, arms control has nothing to grip.
