A single AI-guided drone killed three civilians in Zaporizhzhia, Ukraine, on 6 July 2026. Tetiana Bubynets, a 19-year-old university student, died alongside Oleksiy Svirin, 41, and Roman Karpiy, 48, when the Russian drone struck a wall and exploded. Human operators had programmed it to fly toward a specific gas station. But the machine appears to have chosen its precise target — possibly propane tanks — on its own. Later analysis traced the drone's guidance to an experimental AI system running on a computer chip sold by Nvidia.

That incident, described by Dr Emily Camins, a lecturer in international humanitarian law at the University of Western Australia Law School, sits at the center of a growing legal problem. Lethal autonomous weapons systems, or LAWS, are now widespread in armed conflict. Yet international law offers their victims almost nothing.

What does the law currently say about victims of LAWS?

Almost nothing, according to Camins. The main multilateral forum on the issue — the Group of Governmental Experts (GGE) on Emerging Technologies in the Area of Lethal Autonomous Weapons Systems — has concentrated on accountability, not victim support. Its latest report, CCW/GGE.1/2026/CRP.1/Rev.1 of 3 September 2026, addresses doctrinal questions of accountability for autonomous systems (paras 42-44).

The 5 June 2026 rolling text of the GGE follows the same path as its 12 May 2025 predecessor. It seeks to ensure that States, parties to armed conflict, and individuals remain 'responsible … and accountable' for decisions regarding LAWS (chapter V). Where victims have appeared at all — as in a 2024 working paper submitted by Bulgaria, Denmark, France, Germany, Italy, Luxembourg and Norway (CCW/GGE.1/2024/WP.3) — the framing has been State responsibility for internationally wrongful acts or IHL violations (para 10). Camins writes that reparation or measures expressly supporting victims of LAWS have been 'largely absent from the discussion.'

Why is compensation not the whole answer?

Compensation dominates discussions of reparations for IHL violations, but it is one option among several. The Articles on the Responsibility of States for Internationally Wrongful Acts (ARSIWA) set out three forms: restitution, compensation and satisfaction.

• Restitution (Article 35) requires a State to 're-establish the situation which existed before the wrongful act was committed' — rarely possible where death or personal injury is involved.
• Compensation (Article 36, reflected in Article 91 of Additional Protocol I and customary IHL) covers financially assessable damage from a wrongful act.
• Satisfaction (Article 37) may involve acknowledgement of the violation, a statement of regret, or an apology.

The victim-oriented UN Basic Principles and Guidelines on the Right to a Remedy and Reparation add rehabilitation and guarantees of non-repetition. All of these, however, depend on establishing a violation of IHL or international human rights law in the first place.

That dependency is precisely where autonomous weapons cause trouble. LAWS, Camins argues, are likely to make the path to reparations more challenging by obscuring who made the decision that caused harm. Two questions follow. Where LAWS make erroneous targeting decisions based on opaque data, how can a State meaningfully guarantee non-repetition? And where civilian harm results from opaque machine decision-making, how meaningful can an apology be? The GGE report, with its emphasis on human control and judgment, addresses these questions only partially.

Is there a workable precedent?

Yes, and it comes from humanitarian disarmament law. States have often resisted conferring reparative rights directly on individuals. But they have accepted alternative, victim-centric obligations through weapons treaties.

The 2008 Convention on Cluster Munitions (Article 5) and the 2017 Treaty on the Prohibition of Nuclear Weapons (Article 6) both adopt victim assistance frameworks. These impose primary obligations on States to provide medical care, rehabilitation, psychological support, and economic and social inclusion measures to victims within their jurisdiction or control. The duty extends to family members of those killed.

The critical feature is that these are primary obligations. Victims do not need to prove legal liability for any violation. In the LAWS context, where responsibility may be especially contestable, that structure would offer far more accessible support than measures built strictly on State responsibility principles.

There is also precedent inside the Convention on Certain Conventional Weapons framework itself — the treaty hosting the GGE process. Building on Article 6(3) of the 1997 Anti-Personnel Mine Ban Convention, Protocol V on Explosive Remnants of War (2003) states in Article 8(2) that States parties 'in a position to do so shall provide assistance for the care and rehabilitation and social and economic reintegration of victims of explosive remnants of war.'

What does this mean for practitioners and negotiators?

For lawyers advising States and civil society organizations engaged in the CCW/GGE process, the practical takeaway is specific. A victim assistance provision modeled on Protocol V would not require resolving the harder disputes over human control, accountability, or legality of LAWS that have stalled the negotiations. It would function as a primary obligation, decoupled from liability questions that autonomous decision-making makes harder to answer.

Camins frames it as a modest measure. Further civilian harm from LAWS is likely, she notes, and even lawfully deployed weapons cause death, injury and destruction — harm that is 'awful but lawful.' Experts have warned of targeting errors, discrimination, murky accountability frameworks and desensitization to the effects of armed conflict. To the extent LAWS lower the threshold for armed conflict, a proposition Camins calls contested, the risk of civilian harm grows.

Her closing argument is direct: introducing a victim assistance provision into the GGE process would help ensure that those most likely to be harmed by LAWS are not forgotten as autonomous weaponry reshapes warfare. The three people killed in Zaporizhzhia on 6 July 2026 had no legal avenue designed for them. The question for negotiators is whether the next GGE text will change that.