There is a number that has reshaped the global debate on AI warfare more than any other: 20 seconds.

That is the amount of time human analysts reportedly spent reviewing each target flagged by an Israeli military AI system called Lavender before authorising a strike. Not 20 minutes. Not 20 hours of intelligence cross-referencing. Twenty seconds often, according to the investigation that exposed it, used only to confirm the flagged target was male.

This is not a hypothetical scenario from a defence conference panel. It happened. It is documented. And it has triggered the most consequential argument in modern warfare: when an algorithm decides who to kill, who or what is actually deciding?

What Lavender and Gospel Actually Did

A report by Israeli publications +972 and Local Call disclosed that the IDF had integrated two AI systems Lavender and Habsora, known as “The Gospel” in Gaza. The Gospel was used to select physical strike targets, while Lavender generated human targets, at one point flagging as many as 37,000 potential individuals, predominantly “military-aged males” classified as potential militants.

The systems fed targeting reports to human analysts. If an analyst judged that a flagged target met the criteria, the judgement passed to a more senior intelligence officer for final confirmation of the strike. Human operators have since admitted their role functioned largely as a rubber stamp investing as little as 20 seconds of review before authorising a strike, often only checking that the flagged target was male.

A February 2025 report titled “Deadly Algorithms” confirmed that Lavender could approve targets within 20 seconds, often without substantive human review, and that since October 2023, the system had compiled its list of 37,000 names without independently verifying the military status of those flagged.

The technical sophistication of the system is, in a sense, beside the point. What changed warfare was not the algorithm’s accuracy. It was the speed at which human judgement was bypassed while technically remaining “in the loop.”

The Phrase That Defines the Entire Controversy: “Meaningful Human Control”

The core challenge is that systems like Lavender and Gospel are not “fully autonomous” in the science-fiction sense of a robot independently deciding to kill. They are human-machine hybrids in which human oversight has been compressed to near-zero without being formally removed.

This distinction is the entire legal and ethical battlefield. Supported by the International Committee of the Red Cross and other human rights organisations, the standard of “meaningful human control” asserts that humans must maintain genuine oversight over critical functions including target selection and engagement — to ensure compliance with international humanitarian law.

The 20-second review window demonstrates that meaningful human control can be present in form while being absent in function a gap that policymakers working on autonomous weapons treaties must now grapple with directly.

AI decision support systems can integrate and analyse vast amounts of data from multiple sources in seconds to produce targeting recommendations but their speed and scale, compounded by automation bias, can lead to simple rubber-stamping by the human user, replacing human judgement rather than supporting it.

How Fast the Kill Chain Has Actually Compressed

The acceleration is not limited to one conflict. It has become the operating standard.

The head of US Central Command, Brad Cooper, disclosed that AI systems were being used to compress “processes that used to take hours and sometimes days” into seconds. In a recent operation, the US military conducted 900 strikes in the first 12 hours — confirming that AI integration into operational planning had compressed the kill chain into an operating baseline where military campaigns run at machine speed, with humans approving rather than originating most targeting decisions.

The Ukraine-Russia conflict has become the world’s preeminent testing ground for AI-enabled weapons. Drone strike events worldwide climbed from 364 in 2018 to over 42,000 in 2025, with Ukraine and Russia together accounting for roughly four out of every five strike events globally since 2018. Ukraine’s Deputy Defense Minister has confirmed the country has “partially implemented” autonomous targeting systems in some devices, with an explicit strategic objective of progressing toward fully autonomous targeting and swarm coordination.

The Vote That Exposed the Global Divide

If there were ever a single moment that revealed exactly where the world’s major military powers stand on this question, it happened at the United Nations in November 2025.

The November 2025 UN General Assembly resolution on lethal autonomous weapons was supported by 156 states, arguing that autonomy in weapons systems raises significant humanitarian, legal, security, technological, and ethical challenges by undermining meaningful human oversight in the use of armed force. The resolution was voted against by Russia, Israel, and the United States, while China abstained.

This was not the first such vote. In December 2024, the UN General Assembly had already adopted a resolution on autonomous weapons by 166 votes in favour, 3 against, and 15 abstentions affirming that international humanitarian law applies throughout the life cycle of military AI. At the first dedicated UN meeting on autonomous weapons in May 2025, attended by 96 countries, Secretary-General António Guterres advocated for a legally binding instrument to ban such systems by 2026, describing them as “politically unacceptable and morally repugnant.”

Over 120 countries have now endorsed a new international treaty on autonomous weapons systems. But major-power opposition makes the likelihood of a binding instrument by the 2026 deadline slim.

What World Leaders Are Actually Saying

The language from senior officials has grown unusually direct for diplomatic settings.

Addressing the General Assembly in September 2025, Ukrainian President Volodymyr Zelenskyy warned: “It is only a matter of time before drones are fighting drones, attacking critical infrastructure and targeting people fully autonomous all by themselves … no human involved.”

European Council President António Costa echoed the concern: “Most dangerous, the development of lethal autonomous weapons systems threatens to remove human accountability from decisions of life and death. The risks are real: miscalculation, escalation, and proliferation. We must act before the tipping points become irreversible.”

Guterres framed the stakes starkly at the Global Dialogue on AI Governance: “The question is whether we will govern this transformation together or let it govern us.”

Why a Treaty Keeps Failing to Materialise

The obstacle is not a lack of proposed solutions. It is a lack of political will among the states with the most advanced systems.

Scholars of international humanitarian law have proposed several responses to the rubber-stamp problem: minimum review time standards, mandatory transparency reporting on AI use in targeting, prohibition on AI systems trained on unverified datasets, and criminal accountability standards addressing rubber-stamp approval chains. None of these proposals currently has sufficient state support to advance.

Defense establishments in the United States, United Kingdom, China, France, and Australia have all indicated interest in the capabilities demonstrated in Gaza. The key lesson militaries are drawing is not that the specific systems should be replicated, but that the integration model AI flagging targets, human rubber-stamping, automated logistics represents a viable operational architecture. The controversy around disclosed collateral-damage ratios is treated by most defence establishments as a disclosure problem rather than a capability problem.

China’s PLA has been particularly attentive, with published military AI doctrine emphasising “intelligentized warfare” and calling for AI integration across all phases of the kill chain.

The Moral Argument That Refuses to Go Away

Beyond the legal mechanics, a deeper philosophical objection has gained traction among ethicists and military lawyers alike.

Ethically, the automation of force challenges core principles of human dignity. One of the most pressing areas of ethical concern is the direct targeting of individuals a process that risks depriving humans of their dignity, dehumanising the use of force, and violating the principle of nondiscrimination.

As philosopher Marie-des-Neiges Ruffo de Calabre wrote in Le Monde: “Human judgment must remain central in the decision-making loop for firing for moral, practical and safety reasons. Giving autonomy to the firing decision itself constitutes a moral red line that must not be crossed.”

The market driving this technology forward shows no sign of slowing regardless of the ethical debate. The global autonomous weapons market reached $14.2 billion in 2024 and is projected to grow to $33.47 billion by 2032.

The Question Nobody Has Answered

The most unsettling part of this story is not that an algorithm helped identify targets. Algorithms have assisted human decision-making in warfare for decades. It is that the scale and speed of AI-generated targeting has outpaced the legal and institutional frameworks designed to hold someone accountable when it goes wrong.

As algorithms begin to make decisions that determine who lives and who dies on the battlefield, the rise of AI-driven autonomous weapon systems is forcing a re-examination of the most basic principles of international humanitarian law.

When a strike kills the wrong person, who answers for it the analyst who reviewed it for 20 seconds, the officer who authorised it, the engineer who trained the model, or no one at all? To uphold the rule of law, states and international bodies need to insist on three concrete measures: meaningful human oversight at critical moments, traceable and transparent AI-enabled decisions, and clear rules that hold everyone in the chain developers, commanders, and operators accountable when things go wrong.

As of today, no binding international law requires any of those three things. The systems are already operating. The treaty to govern them is still, by every account, years away.

📌 Read Also:

AIWala News