AINEWS
Both sides of the Ukraine war now field drones whose onboard sensors pick and engage targets on their own — with no human in the loop — and the three places where 'the test for whether a human still has to approve each strike' could still become a
Sensor-selected targeting is not coming. It is in operational use on both sides of the Russia–Ukraine war, where uncrewed aerial vehicles already account for 70–80% of casualties in the current attritional phase. The drones look similar from the air. The legal and moral questions they raise are not the same.
Two engineering responses to Russian radio-frequency jamming now share the same battlefield. The first is the fibre-optic-tethered drone: a wound-out cable replaces the radio link, so the operator keeps full control and the drone keeps flying. The second is sensor-selected targeting, the class the international community labels "lethal autonomous weapons systems" (LAWS): once launched, the system picks and engages targets on its own, with no human in the loop. Both are answers to the same problem. One is a workaround with an environmental footprint. The other is a workload-shedder with an accountability gap.
The accountability gap is the part that policy has not caught up with. When a sensor-selected system kills, the chain breaks at every human station. The operator did not select the target. The developer wrote a model, not an order. The manufacturer shipped a platform, not a strike. The commander authorised a deployment envelope, not a person. [Human Rights Watch's 2012 Losing Humanity report](https://www.hrw.org/report/2012/11/19/losing-humanity/case-against-killer-robots) named the problem first and called for a preemptive ban on exactly this ground: there is no one left in the loop to hold responsible. More than a decade later, HRW's April 2025 brief is still making the same argument, because the systems it warned about are now being fielded, not forecast.
There is no agreed international definition of LAWS. The working characterisation used by the UN and most major organisations — weapons that, once activated, identify and apply force to targets on the basis of sensor processing rather than human input — is exactly that: a working characterisation, not a treaty text. PassBlue reported in May 2025 that diplomatic activity around a possible LAWS treaty was still alive but stumbling, and the EU's October 2025 statement to the UN General Assembly First Committee on conventional-arms governance shows Brussels actively weighing in, rather than waiting for Washington or Beijing to move first.
Fibre-optic-tethered drones are a separate, smaller problem that is not separate enough. The cable defeats jamming and keeps the operator in command, which is the point. It also drags kilometres of plastic and glass through fields and forests, with a footprint that lasts well past the war. A Latvian report covered by Politico puts three in four drone strikes in the "mass casualty" category on the Ukrainian side, using a different denominator than the 70–80% battlefield-wide figure. The two numbers should not be merged, but the direction is the same: the attritional drone war is now the dominant way this war produces its dead and wounded.
Three pressure points decide whether "meaningful human control" becomes a binding rule or a slogan. First, the UN Convention on Certain Conventional Weapons (CCW) process in Geneva, where the LAWS working characterisation lives and where any treaty would have to be negotiated. Second, national autonomous-weapons policy, where the US, UK, France, and others have issued internal directives that stop short of a prohibition and that diverge in the kind of human review they require. Third, arms export rules, where the EU's existing common position already restricts exports of weapons used for repression but does not yet treat sensor-selected targeting as a category. France has been reported to be pushing back against tighter EU AI rules in the same period, which is why the EU's October UNGA statement matters more than its usual boilerplate.
The Russia–Ukraine war is not the first war to use uncrewed systems. It is the first in which the dominant casualty mechanism is a system that, for a growing share of its strikes, decides for itself who to kill. Sensor-selected targeting is not error-free. It is differently errored: miscalibrated sensors, target-library drift, sign errors in models trained on limited data. The legal vacuum around it is not a future-tense problem. It is the working condition under which the next several years of drone war will be fought, and the Geneva process is the only one currently structured to write the rule before the rule is set by repeated use.