A lethal autonomous weapon is a weapon system that, once switched on, can search for, select, and fire on a target without a human approving that specific strike. The idea alarms military ethicists and diplomats alike, and the United Nations has spent more than a decade trying — and failing — to agree on a treaty that would restrict or ban them.
What makes a weapon “autonomous”
Most weapons today, including armed drones, are remotely piloted: a human somewhere watches a video feed and decides when to fire. An autonomous weapon removes that person from the final decision. Militaries describe this with a rough spectrum: “human-in-the-loop” systems need a person’s go-ahead for every strike; “human-on-the-loop” systems let a human supervise and abort a machine’s decision; “human-out-of-the-loop” systems act on their own once activated.
Some autonomous functions are old and uncontroversial — ship-defense systems like the US Navy’s Phalanx have shot down incoming missiles without human input since the 1970s, because the threat moves too fast for a person to react. What’s new is autonomy applied offensively, against people. A 2021 UN Panel of Experts report described a Turkish-made loitering munition, the STM Kargu-2, as having potentially hunted retreating fighters in Libya without a human directing each attack — one of the first documented cases of its kind. Since then, cheaper sensors and AI-based targeting software have made autonomous targeting far easier to build, and multiple militaries are now testing drones and drone swarms with reduced human oversight.
Why the UN wants a ban
UN Secretary-General António Guterres has called such weapons “politically unacceptable” and “morally repugnant,” arguing that decisions to take a human life should remain “forever human.” The UN and the International Committee of the Red Cross have jointly urged states to adopt a legally binding instrument, citing both the ethical problem of delegating lethal force to a machine and the practical one of who is accountable when an autonomous system kills the wrong person.
Why there’s still no treaty
Diplomats have discussed lethal autonomous weapons since 2014 inside the Convention on Certain Conventional Weapons (CCW), the main UN forum for restricting weapons considered excessively injurious. The CCW works by consensus, meaning any single member can block a proposal — and a handful of states with major military AI programs, including the United States, Russia, India, and Israel, have repeatedly done so, arguing that existing international humanitarian law is sufficient and that a binding ban is premature or unenforceable.
Frustrated by that gridlock, the UN General Assembly took up the issue directly. In December 2024, 166 countries voted in favor of a resolution moving the debate to the General Assembly, where consensus isn’t required; only 3 voted against, with 15 abstentions. The first General Assembly meeting dedicated to the topic followed in May 2025, drawing delegations from 96 countries. More than 120 states now say they want to negotiate a binding treaty, but the states building the most advanced autonomous systems have not joined that call, so formal treaty negotiations still haven’t begun — only consultations.
In the news
The standoff was on display again this week, when the UN convened its first Global Dialogue on AI Governance, bringing all 193 member states to Geneva. Guterres used the gathering to renew his call for a treaty. Georgia’s president also addressed the dialogue, warning that concentrated control over AI must not become “a weapon of totalitarian control” — a broader worry about unaccountable AI power that autonomous weapons put in its starkest form.
Why it matters for Georgia
Georgia does not build autonomous weapons, but it isn’t a bystander to this debate: its president used the UN’s inaugural AI governance dialogue to argue publicly for binding international rules on how AI-enabled power, including military power, gets used and checked. For a small state, a rules-based order — rather than one where a handful of major militaries set the norms unilaterally — is generally the safer bet.
FAQ
Are autonomous weapons already in use? Partially. Fully autonomous defensive systems have existed for decades, and there are credible reports of autonomous or near-autonomous offensive use in recent conflicts, but no state has officially confirmed deploying a weapon that selects and kills human targets without any human approval.
Is there any international law that already applies? Yes — existing international humanitarian law (rules on distinction, proportionality, and accountability) applies to all weapons, autonomous or not. The dispute is over whether that’s enough, or whether a new, autonomous-weapons-specific treaty is needed.
Why can’t the UN just vote to ban them? The traditional forum for weapons treaties, the CCW, requires consensus, so any one opposed state can block a binding rule. Moving the debate to the General Assembly sidesteps that veto, but a General Assembly resolution carries less legal weight than a ratified treaty.
Sources: UN News coverage of the Secretary-General’s remarks, Human Rights Watch reporting on the 2024–2025 General Assembly process, the ICRC/UN joint call on autonomous weapons, and the Arms Control Association.