PROFILE: Autonomous Drones and the Legal Void — Who Is Responsible When a Machine Kills?

Introduction: The Silent Revolution Changing Warfare

$55 Billion and a Question Without an Answer

The global defense drone market will reach, according to analyses by BriefGlance and several specialized firms, over $55 billion by 2032 — some projections even advance the figure of $90 billion by the end of the decade. The Pentagon included in its budget request for 2027 over $54 billion for its "Drone Dominance Program" — aiming to deploy "hundreds of thousands of platforms." This technological revolution, rapid and massive, has outpaced the ethical, legal and political reflection that should frame it. This profile is not of a person. It is of a technology that, without regulation, risks rewriting the rules of war in ways nobody yet fully measures.

The Question That Underpins Everything Else

When an autonomous weapons system decides on its own — without direct human validation — to fire on a target and kills civilians by error, who is responsible? The manufacturer who programmed the algorithm? The military commander who defined the engagement parameters? The state that bought the system? The soldier who deployed it? This question, seemingly abstract, is the central question of robotic warfare. It does not yet have a clear international legal answer — and this absence of an answer is itself a political choice.

The Market: $55 Billion and a Race Nobody Wants to Lose

Industrial Players and Their Flagship Systems

The Eurosatory 2026 exhibition illustrated the acceleration of this industry: the company VisionWave was presenting its TALON™, D-FLY™ and STRATUM™ systems — platforms designed for reconnaissance, light attack and adversary defense saturation missions. American company Ondas Holdings was announcing over $150 million in orders in the second quarter of 2026 for its industrial security drones convertible to defense applications. These two examples illustrate a broader phenomenon: the boundary between civilian and military drone markets is progressively dissolving.

The AUKUS Dynamic and the Autonomous Swarm

The Army Warfighting Experiment 2026 at Salisbury Plain — where up to six autonomous drones were coordinated from a tripartite command hub — illustrates the operational direction: drone swarms capable of acting in a coordinated way with minimal human intervention are the development priority of the main Western armed forces. These swarms raise specific questions about delegation of lethal authority: when a distributed algorithm decides the target and moment of engagement, which human has "pulled the trigger" in the legal sense?

The answer from the U.S. Department of Defense Directive 3000.09 is that humans must retain "full responsibility" over lethal decisions. But this deliberately vague formulation does not specify what this responsibility means when the system acts in a fraction of a second, well beyond human supervision capacity in real time.

The International Legal Void: Geneva 2026 Without a Treaty

The GGE LAWS: Ten Years of Dialogue Without Agreement

The UN Group of Governmental Experts on Lethal Autonomous Weapons Systems (GGE LAWS) has been holding sessions in Geneva since 2014. Its March 2026 session ended without substantial progress toward a binding treaty. Positions are locked: one group of countries (including Austria, New Zealand, Mexico) advocates for a treaty prohibiting LAWS without direct human supervision. Another group (including the United States, Russia, China, Israel) refuses any fixed technical threshold, arguing that autonomy thresholds are too context-dependent to be defined a priori.

Washington Refuses Fixed Thresholds — and the Reason Is Strategic

The American refusal to set technical thresholds in international treaties is motivated by coherent strategic logic: imposing limits on drone autonomy would freeze a technological advantage that evolves rapidly. The United States currently has the advantage in autonomous systems — it does not want that advantage erased by an international rule that would apply equally to it and to China. This is selective multilateralism — the same observed in commercial or nuclear discussions.

China and Russia have adopted similar positions for symmetrical reasons: they refuse international rules that would limit their investments in areas where they seek to close their technological gap. The result is a normative vacuum maintained by consensus of inaction — all major actors prefer the absence of rules to rules that would constrain their own programs.

Operational Gray Areas: Ukraine, Gaza, South China Sea

Ukrainian Drones: Between Autonomy and Supervision

In Ukraine, thousands of drones are deployed each week on both sides of the front. Ukrainian drones — notably models like the Ukrainian reversed Shahed-136 (an improved copy of the captured Iranian drone) and domestic models of the FPV (First Person View) type — operate with varying degrees of autonomy depending on the mission. AI-based image recognition navigation systems allow some of these drones to identify and attack targets without a human pilot in direct contact. Whether these systems constitute "LAWS" in the GGE sense remains deliberately ambiguous.

Gaza and Allegations About Targeting Algorithms

In Gaza, allegations documented by researchers from Fordham University and the magazine +972 suggest that the Israel Defense Forces (IDF) use AI-assisted targeting systems — notably the "Lavender" system — to generate lists of potential Hamas targets, with minimal human validation process. Israel has neither confirmed nor denied the use of these systems. If these allegations are accurate, they illustrate precisely the problem the GGE LAWS seeks to regulate: autonomous systems that influence lethal decisions in a high civilian density conflict.

The difficulty is that without a treaty defining what constitutes a prohibited "autonomous weapons system," there is no international legal basis to challenge the use of these systems — even if their deployment produces results that would shock the international conscience.

Non-State Actors and the Democratization of Lethality

Hezbollah, Iran and the Proliferation of Autonomous Drones

Autonomous drone systems are no longer the exclusive preserve of major military powers. Hezbollah — backed by Iran — deployed during the 2023-2024 fighting in Lebanon drones with autonomous terminal guidance capable of targeting specific Israeli military vehicles. The Houthis in Yemen used long-range cruise drones to strike commercial vessels in the Red Sea — a capability that, a decade ago, existed only in the arsenals of first-tier powers.

Uncontrolled Proliferation and Its Consequences

The technological democratization of lethal autonomous drones means that any actor — state or non-state — with a modest budget can now acquire or manufacture systems capable of autonomous attacks. Conversion kits to transform commercial drones into military platforms are available on online gray markets. This proliferation makes international regulation even more urgent — and even more difficult. Regulating states would already be complex. Regulating non-state actors operating outside any international legal framework is a task that exceeds current international law tools.

The Geneva Commission on Drone Proliferation, created in 2025, has identified over 100 non-state actors with access to lethal-capable drone systems. This figure, continuously rising, illustrates the pace at which technology outstrips regulatory frameworks.

Companies Facing Their Responsibilities: The Ethics of Military AI

Google, Microsoft and the Dilemmas of Engineers

In 2018, thousands of Google employees had signed a petition against Project Maven — a Pentagon contract to develop image analysis algorithms intended to improve drone targeting. Google ultimately declined to renew this contract. In 2026, this internal resistance is far less visible: defense budgets are substantial, competition with China is presented as a national imperative, and employees who oppose military contracts face growing institutional pressures.

Responsible AI Principles — Between Declarations and Practices

Microsoft, Amazon and Google have published "responsible AI principles" that display their commitment to ethical and safe systems. These principles, carefully drafted, contain formulations sufficiently vague to allow most of the military applications their authors claim to want to avoid. Independent verification of whether these principles comply with actual practices is practically impossible — military contracts are classified, targeting algorithms are industrial and national secrets.

This opacity is the breeding ground of structural irresponsibility: nobody can verify that the principles are respected, so nobody can establish the violation, so nobody is formally responsible. The circle is closed — and civilians killed by algorithmic errors have no one to turn to.

Possible Solutions: What Do Experts Propose?

Three Approaches to Regulating LAWS

International humanitarian law experts generally propose three approaches to regulating lethal autonomous weapons systems. The first — championed by humanitarian organizations and some states — is an outright prohibition on LAWS without direct human oversight in the targeting cycle. The second is a certification framework: systems could only be deployed if certified as compliant with minimum standards of distinction between combatants and civilians. The third — favored by major powers — is a voluntary code of conduct without binding enforcement mechanism.

The International Committee of the Red Cross Proposal

The International Committee of the Red Cross (ICRC) proposed in 2024-2026 a compromise position: requiring that any autonomous system with lethal capability be subject to meaningful human control in contexts where it risks violating international humanitarian law — notably the distinction between civilians and combatants and proportionality. This formulation of "meaningful supervision," more operational than an absolute prohibition, could form the basis of a negotiated agreement.

The difficulty is that "meaningful" remains an open term subject to interpretation. The same term could cover equally an officer who examines each target for ten minutes or an operator who approves a list of a hundred targets in five minutes. Without a precise technical definition, this framework risks becoming as vague as the companies' "responsible AI principles."

Ukraine as an Involuntary Laboratory

What the War Teaches About Drone Autonomy in Real Conditions

The war in Ukraine has become despite itself the largest laboratory of autonomous drones in real conditions in history. Thousands of drones on both sides of the front have been deployed in dense electronic warfare environments, with active countermeasures, in adverse weather conditions and in areas with high civilian density. The operational data collected — on targeting error rates, electronic countermeasure effects, algorithm performance in constrained environments — is of inestimable value for military planners.

Lessons Passed On to Beijing and Moscow

These lessons do not only benefit Ukrainian forces and their Western allies. Russia is integrating them into its drone development programs. China, closely watching Ukrainian operations, is accelerating its own autonomous combat drone programs drawing lessons from errors and successes observed in Ukraine. The Ukraine war is accelerating the autonomous arms race worldwide — a consequence that nobody had fully anticipated when the first Geneva discussions on LAWS took place in 2014.

For Ukraine, this reality is ambiguous: its drones heroically defend it today. But the systems it develops and tests feed a global market and a proliferation dynamic that could threaten it tomorrow in different geopolitical configurations. Zelensky knows that technological superiority in drones is temporary — it must be accompanied by an international governance architecture to be durable.

Conclusion: Responsibility Is Not Optional

The Legal Void as a Political Choice

The absence of a binding international legal framework on lethal autonomous weapons systems is not an oversight or a technical gap. It is a deliberate political choice by major military powers that prefer to maintain their short-term technological freedom of action at the expense of a long-term collective security framework. This choice has a cost: civilians killed by algorithmic errors have nobody to pursue, soldiers exposed to autonomous attacks have no specific legal protection, and proliferation to non-state actors continues unhindered.

The Responsibility Question Will Not Wait for Politicians

Technology advances. Autonomous drones deploy. The war in Ukraine and operations in Gaza demonstrate that high-autonomy systems in lethal decisions are already an operational reality. The responsibility question will not wait for an international political decision — it will be raised by families of killed civilians, by national courts, by ICC bodies. And when it is raised, the absence of a clear normative framework will transform each incident into an unprecedented diplomatic crisis. Building this framework now, while minds are focused, is a moral and strategic obligation. Not later. Now.

Signed Maxime Marquette, columnist

Columnist's transparency box

Method

This profile of a technology draws on open sources: market reports (BriefGlance, defense analysis firms), official UN GGE LAWS documents, ICRC publications, Pentagon reports, and academic analyses on LAWS legal liability. Allegations about the "Lavender" system are based on investigations published by verified investigative media (+972 Magazine). Israel not having confirmed this information, it is presented as documented allegations and not established facts.

Positioning

The columnist supports the development of autonomous defense capabilities for democracies facing authoritarian threats, while advocating for a binding international legal framework that establishes rules of engagement, accountability mechanisms and proliferation limits.

Sources

Primary sources

BriefGlance — The New Arms Race: Autonomous Drones Fuel a $55B Defense Revolution — 2026

UN GGE LAWS — Session Reports March 2026

ICRC — Autonomous Weapons Systems: What Constitutes Meaningful Human Control — 2026

Secondary sources

Human Rights Watch — Autonomous Weapons Systems — 2026

Campaign to Stop Killer Robots — State of Regulation — 2026

Foreign Policy — Autonomous Drones: The Accountability Gap — June 2026