COLUMN: AI enters the kill chain — and nobody is truly ready for that

Introduction: A war that changes in nature

June 25, 2026: a disclosure that disturbs

On June 25, 2026, Business Insider publishes an analysis by Ukrainian analyst Serhii "Flash" Beskrestnov on the next evolution of Shahed drones: the future version of the Iranian suicide drone used extensively by Russia against Ukraine would allow artificial intelligence to choose its target automatically, without human intervention. Simultaneously, Euromaidan Press reveals that Ukraine's A1 Centre, created in March 2026, already integrates AI into the kill chain — the decision sequence leading to a lethal strike.

These two pieces of information, taken together, mark a turning point in the history of warfare. For the first time in a high-intensity conflict, artificial intelligence is no longer merely assisting human decision-making — it is making or directing lethal decisions in real time. This is a threshold that philosophers, lawyers, and military thinkers had been debating in theory for years. It has just been crossed.

What this column means to say

I am not going to pretend to have all the answers about the implications of AI in warfare. That would be dishonest. These issues — ethical, strategic, legal, operational — exceed the analytical capacity of any single columnist. But I can document what is happening, reflect on what it means, and share the questions that have haunted me since I read this information.

This column is a journey through the available data — Ukraine, China, Russia, the United States — and through the questions they raise. No definitive answers. Just the honesty of an observer who finds that the world has just crossed a threshold whose consequences are not yet measurable.

Ukraine and the A1 Centre: AI in the kill chain

A centre created in urgency, but with a vision

Ukraine's A1 Centre, created in March 2026, is Ukraine's institutional response to the technological revolution underway on the battlefield. According to Euromaidan Press (June 25, 2026), this centre integrates AI into the kill chain at several levels: terminal drone guidance in the final seconds before impact, robotic turrets capable of deciding to engage a target, and management of drone swarms.

Terminal AI guidance is particularly significant. In the final seconds of a drone flight, the human operator often loses connection with the device — enemy jamming, distance, atmospheric conditions. AI can take over to identify and engage the final target. The human decided on the launch and the general target. But the final lethal decision belongs to the algorithm.

The Ukrainian confirmation: 10 autonomous drones, 10 Russian soldiers killed

According to DroneSense (June 22, 2026), Ukraine confirmed a shocking disclosure: 10 autonomous drones killed Russian soldiers near Bakhmut without human intervention. This is the first public confirmation of a lethal autonomous operation of this nature in a high-intensity conflict. These drones identified, targeted, and eliminated human beings without an operator validating each engagement.

This fact is of considerable strategic and ethical importance. It confirms that Lethal Autonomous Weapons Systems (LAWS) have moved from concept to operational reality. The academic debate over their legitimacy has now been overtaken by facts on the Ukrainian battlefield.

Ukraine prepares: 25,000 UGVs and a dedicated battalion

Unmanned ground vehicles

Ukraine's commitment to military robotics is not limited to the air. Ukraine has created the first dedicated battalion for unmanned ground vehicles (UGVs). And according to available data, 25,000 UGVs were ordered in the first half of 2026 alone.

These autonomous or semi-autonomous ground vehicles can carry munitions to front lines, evacuate casualties under fire, or participate directly in offensive operations. Combined with aerial drones, they create a robotised combat ecosystem that reduces exposure of human soldiers to the most dangerous situations — and increases the possible intensity and duration of operations.

The race to swarms

Drone swarms — swarms of dozens or hundreds of drones coordinated by AI — represent the most disruptive evolution. Manually coordinating 10 drones simultaneously is already complex. Coordinating 100 or 1,000, in real time, with millisecond decision-making — that is humanly impossible. Only AI can manage this complexity.

Ukraine is investing in this capability — the A1 Centre explicitly works on "pilot swarms." This is a race against Russia, which is developing its own swarms, and against China, whose capabilities in this domain are the most advanced in the world. Whoever masters autonomous swarms first will hold a potentially decisive strategic advantage.

Russia: Lightning-13 and the AI-enabled Shahed drones

1,400 jet drones since early 2026

Russia is not standing still in this technological race. According to Ronin's Grips (June 20, 2026), Russia is deploying the "Lightning-13" (Molniya-2) system — a high-speed jet or electric drone. Since the start of 2026, Russia has launched 1,400 units of this system — compared to 180 launched across all of 2025. That is an eightfold increase in production and deployment pace.

This acceleration is alarming. It demonstrates that despite sanctions, electronic component shortages, and documented infrastructure damage, Russia is maintaining and increasing its drone production capacity. It is adapting — with alternative components, local substitutions, alternative supply channels through Iran, China, and North Korea.

Shahed drones with autonomous targeting AI

The most troubling revelation comes from analyst Beskrestnov: the next generation of Shahed — the Iranian suicide drone used extensively by Russia — will integrate AI capable of choosing its target automatically. No operator needed to designate the final target. The drone analyses the environment, identifies objects according to predefined criteria, and engages.

This evolution makes current jamming systems far less effective. Many Ukrainian and allied anti-drone systems work by cutting the communication link between the drone and its operator. An autonomous AI drone that no longer needs a communication link for final decisions is structurally resistant to this form of countermeasure.

China: Atlas and 96 drones per operator

CETC March 2026: the demonstration

In March 2026, CETC (China Electronics Technology Group Corporation) presented the Atlas system: a device allowing a single operator to direct 96 drones equipped with "individual AI intelligence" for coordinated precision strikes. This demonstration is documented by Chosun Biz (June 22, 2026).

The concept of "mass with precision" — sending dozens or hundreds of intelligent drones simultaneously — represents a break from classical military doctrines. A single human operator supervises 96 autonomous lethal agents. The human engagement rule is reduced to its simplest expression: deciding to launch the swarm, without individual control of each strike.

Legal implications: compliance by design

The Lieber Institute at West Point (June 24, 2026) published an analysis on "compliance by design" in Chinese autonomous weapons systems. The argument: if the AI is programmed to respect the rules of war (distinction between combatants and civilians, proportionality, precaution) at a reliability level superior to a stress-impaired human decision on the battlefield, LAWS could be more compliant with international humanitarian law than human soldiers.

This is a provocative but serious thesis. It inverts the usual moral presumption — that only a human can bear legal responsibility for a lethal decision. If an AI is more reliable than a human in respecting discrimination and proportionality protocols, should we maintain the requirement for human control on philosophical grounds, even at the cost of more civilian casualties?

Venture capital and the race for autonomous weapons

12.3 billion USD in defence-tech in H1 2026

Global venture capital in defence-tech reached 12.3 billion USD in the first half of 2026 — almost double the entire 2025 total (9.95 billion USD). This explosive figure reflects the influx of private capital into military technologies linked to AI — autonomous drones, AI-assisted targeting systems, electronic warfare, military cybersecurity.

This transfer of capital from the civilian to the military sector is itself a transformation indicator. The best AI engineers — the most creative, the most ambitious — are now working on military applications, attracted by unprecedented funding and by the sense of contributing to existential stakes. Silicon Valley is militarising its AI.

The race and its proliferation risks

12.3 billion USD invested in H1 2026 means hundreds of startups are developing military AI technologies in several dozen countries. Unlike nuclear weapons, which require heavy and expensive infrastructure, AI-powered autonomous drones can be developed in garages with commercial components.

This technological proliferation risk is one of the most worrying aspects of the autonomous weapons revolution. Today, the main actors are Ukraine, Russia, China, and the United States. In five years, this technology could be accessible to non-state actors, terrorist groups, rogue states — with unpredictable consequences for international security.

International law facing the autonomous revolution

An outdated legal framework

International humanitarian law — the Geneva Conventions, Additional Protocols, rules of war — was designed for human combatants making human decisions. It rests on concepts such as intent, individual responsibility, distinction between combatants and civilians, and proportionality.

These concepts are profoundly ill-suited to autonomous systems. How does one attribute criminal responsibility when an algorithm kills a civilian? To the engineer who programmed the system? To the commander who decided to deploy it? To the NCO who pressed "launch"? The chain of responsibility is broken by systems' autonomy — and no one has yet proposed a legally satisfactory solution.

CISA BOD 26-04 and American AI alerts

In the United States, CISA BOD 26-04 (June 10, 2026) acknowledges that AI can "weaponise" vulnerabilities in hours, shortening remediation timelines to a point where human defences can no longer keep pace. This is an adaptation for cybersecurity — but the logic applies to physical combat.

When decision and action cycles become sub-human in their speed, traditional military doctrine — which assumes humans deliberate and decide — becomes obsolete. The future battlefield is one of algorithmic-speed warfare, where humans supervise ensembles they no longer individually control.

Rules of engagement for autonomous systems: an urgent legal void

International humanitarian law facing lethal autonomy

International humanitarian law — the Geneva Conventions, Additional Protocols — was designed for human combatants making human decisions. It rests on concepts such as intent, individual responsibility, distinction between combatants and civilians, and proportionality. These concepts are profoundly ill-suited to autonomous weapons systems that decide without deliberating.

How does one attribute criminal responsibility when an algorithm kills a civilian? To the engineer who programmed the system? To the commander who decided to deploy it? To the NCO who pressed "launch"? The chain of responsibility is broken by the autonomy of systems — and no one has yet proposed a legally satisfactory solution that commands consensus in international institutions.

CISA BOD 26-04 and American alerts on lethal AI

America's CISA BOD 26-04 of June 10, 2026 acknowledges that AI can "weaponise" vulnerabilities in hours. This is an adaptation for cybersecurity — but the logic applies to physical combat too. When decision and action cycles become sub-human in their speed, traditional military doctrine — which assumes humans deliberate and decide — becomes obsolete.

The future battlefield is one of algorithmic-speed warfare, where humans supervise ensembles they no longer individually control engagement by engagement. The 12.3 billion USD invested in venture capital in defence-tech in H1 2026 are accelerating this transition far beyond what current legal and ethical frameworks can absorb.

Conclusion: The threshold we crossed without a vote

The decision nobody made collectively

No parliament, no international assembly, no referendum decided that humanity would authorise machines to kill human beings autonomously. This decision was made by engineers, military officers, and wartime decision-makers — in urgency, in necessity, in the implacable logic of survival.

Ukraine uses AI in the kill chain because it is fighting for its existence. Russia deploys AI-enabled Shaheds because it wants to destroy Ukraine. China tests Atlas because it is planning long-term military superiority. None of these actors asked humanity's permission. The threshold of lethal autonomy was crossed in the fog of war.

What Zelensky embodies in this debate

For Zelensky, AI in the kill chain is not a philosophical question — it is a question of survival. If autonomous Ukrainian drones can neutralise Russian positions with fewer risks to Ukrainian soldiers, he will use them. His duty to his people takes precedence over the ethical debates those of us in peaceful countries have the luxury of conducting from our offices.

And yet it is precisely from those peaceful offices that the rules must be thought through — not in the fog of war. Ukraine fights with the available tools. It falls to the rest of the democratic world to think through the rules of engagement for autonomous systems before the next war — larger, more automated — begins.

Signed Maxime Marquette, columnist

Columnist's transparency box

Sources and data

This column rests on the dated sources listed below. The figures cited — 25,000 UGVs, 1,400 Lightning-13 units, 96 Atlas drones, 12.3 billion USD in VC, 10 autonomous drones — come directly from the sources without modification. The creation date of the Ukrainian A1 Centre (March 2026) and its mission are documented by Euromaidan Press.

The Ukrainian "disclosure" on the 10 autonomous drones and soldiers killed is reported by DroneSense (June 22, 2026) — the columnist presents it as such, without being able to authenticate it independently. Journalistic caution is appropriate for this type of operational revelation.

Position and limits

The columnist does not claim expertise in autonomous weapons systems, international humanitarian law, or military artificial intelligence. His column is that of a thoughtful observer, not a technical specialist. The ethical questions raised reflect his personal doubts — not definitive conclusions on issues that even experts cannot resolve.

He supports Ukraine and acknowledges the legitimacy of using all available technologies in a defensive war. He is also concerned about the long-term implications of normalising lethal autonomous weapons systems. Both positions coexist in this column — the tension between them is acknowledged.

Sources

Primary sources

Business Insider — AI-enabled Shaheds will choose their target automatically, according to Beskrestnov — June 25, 2026

Euromaidan Press — Ukraine wants an AI-driven army, its new defence centre is already putting AI inside the kill chain — June 25, 2026

Secondary sources

Lieber Institute West Point — Compliance by design: China's autonomous weapons development — June 24, 2026

DroneSense — Weekly intelligence brief, 10 autonomous drones Ukraine — June 22, 2026

Chosun Biz — Defence venture capital 12.3 billion USD H1 2026 — June 22, 2026

Ronin's Grips — Military drones SITREP, Russian Lightning-13 — June 20, 2026