The Regulation of Autonomous Weapons: Lethal AI, UN Negotiations

The Technology: What Autonomous Weapons Can Already Do

Lethal Autonomous Weapons Systems (LAWS) are weapons that can select and engage targets without direct human intervention. The technology exists along a spectrum of autonomy. At one end are remotely piloted systems like the MQ-9 Reaper drone, where a human operator makes every engagement decision. At the other end are fully autonomous systems that identify, track, select, and engage targets based on pre-programmed criteria without any human in the loop. The boundary between these categories is where the ethical and regulatory debate concentrates.

Current autonomous and semi-autonomous weapons include The Harpy and Harop loitering munitions — drones that orbit an area and autonomously attack radar emitters. Turkey’s STM Kargu-2 is a quadcopter drone designed for autonomous attack against human targets using facial recognition and machine learning; a UN Panel of Experts report (S/2021/229) noted that Kargu-2 drones were described as lethal autonomous weapons systems that had been programmed to attack targets without requiring data connectivity between the operator and the munition during the 2020 Libyan civil war — though experts have debated whether the system was operating fully autonomously at the time of engagement. South Korea’s SGR-A1 sentry robot, deployed along the DMZ, can detect, track, and theoretically fire on intruders autonomously, though it currently operates in supervised mode. Russia’s Uran-9 unmanned combat ground vehicle saw deployment in Syria.

The naval domain is advancing rapidly. The US Navy’s Sea Hunter unmanned surface vessel can operate autonomously for up to 90 days, traversing up to 10,000 nautical miles without crew. China has been testing advanced drone swarms — including the Jiu Tian drone mothership, reportedly capable of releasing 100 to 150 loitering munitions from internal bays — while integrating AI models like DeepSeek into military decision-making systems. The PLA has tested autonomous swarm technology for both UAVs and unmanned surface vessels, with researchers advocating for minimal human intervention in combat decision-making. The convergence of drone miniaturization, computer vision, edge computing, and reinforcement learning means that autonomous weapons are becoming smaller, cheaper, more capable, and more accessible — including to non-state actors.

The Ethical Framework: Meaningful Human Control

The central ethical concept in the autonomous weapons debate is “meaningful human control” (MHC) — the principle that decisions to use lethal force must involve sufficient human judgment and oversight. The International Committee of the Red Cross (ICRC) argues that meaningful human control requires human commanders to have adequate information about the weapon’s functioning, the target, and the environment; sufficient time to make a considered decision; the technical ability to intervene or abort; and accountability for the consequences.

The ethical arguments against fully autonomous weapons are substantial. First, the principle of distinction — a foundational requirement of international humanitarian law (IHL) — requires combatants to distinguish between military targets and civilians. Current AI systems cannot reliably make these distinctions in the chaotic, ambiguous environments of armed conflict. A child carrying a stick looks different from a soldier carrying a rifle to a human, but may not to a computer vision system trained on limited data. Second, the principle of proportionality requires weighing expected military advantage against anticipated civilian harm — a contextual, value-laden judgment that algorithms cannot meaningfully perform.

Third, there is the accountability gap. When a human soldier commits a war crime, legal accountability is clear: the soldier, their commander, and potentially political leaders can be prosecuted. When an autonomous weapon kills civilians, who is responsible? The programmer who wrote the targeting algorithm? The commander who deployed the system? The manufacturer? The political leader who approved procurement? International humanitarian law assumes human decision-makers; fully autonomous weapons create accountability vacuums that undermine the entire framework of laws of armed conflict. The ICRC has called for new legally binding rules that would prohibit unpredictable autonomous weapons and those designed to apply force against persons, while placing strict restrictions on all others.

The Diplomatic Landscape: From CCW Stalemate to UNGA Momentum

The primary international forum for autonomous weapons regulation has been the UN Convention on Certain Conventional Weapons (CCW), which began hosting informal expert meetings on LAWS in 2014 and formally established a Group of Governmental Experts (GGE) in 2016. Over a decade of discussions later, the GGE has produced no binding instrument. The reasons are structural: the CCW operates by consensus, meaning any single state can block progress, and several major military powers — the US, Russia, India, Australia, and South Korea — have consistently opposed or slowed movement toward a legally binding treaty.

The positions are well-defined. Approximately 30 countries have called for a preemptive ban on fully autonomous weapons — a prohibition before the technology is widely deployed, similar to the Blinding Lasers Protocol (CCW Protocol IV) that banned laser weapons designed to cause permanent blindness before they were used in combat. More broadly, over 120 countries support negotiating a treaty that would prohibit and regulate autonomous weapons systems. The Campaign to Stop Killer Robots, a coalition of more than 270 NGOs working in over 70 countries, actively advocates for this position with significant public support — polls consistently show around 61-62% of citizens in surveyed countries oppose autonomous weapons.

The opposing bloc argues that existing IHL is sufficient, that a ban would be unenforceable and would disadvantage compliant states, and that autonomous systems could actually reduce civilian casualties by being more precise and less emotionally reactive than human soldiers. The US Department of Defense Directive 3000.09 — updated in January 2023 for the first time since 2012 — requires that autonomous weapons allow commanders and operators to exercise “appropriate levels of human judgment over the use of force,” and that development align with DoD AI Ethical Principles, but it does not prohibit the development of autonomous weapons. China has expressed support for a ban on “use” but not on “development” — a distinction that would allow continued weapons research while nominally supporting regulation. Russia has opposed any legally binding instrument.

Frustrated by CCW paralysis, states have shifted to the UN General Assembly, where decisions are made by majority vote rather than consensus. In November 2023, the UNGA First Committee adopted its first-ever resolution on autonomous weapons, with 164 votes in favor. In December 2024, the General Assembly adopted a stronger follow-up resolution with 166 votes in favor and only 3 opposed (Belarus, North Korea, and Russia), creating a new process for open informal consultations on autonomous weapons in New York. A third resolution followed in November 2025 with 156 states in support. In May 2025, UN Secretary-General Antonio Guterres called autonomous weapons “politically unacceptable, morally repugnant,” demanding the conclusion of a legally binding instrument by 2026 — a call issued jointly with the ICRC President.

The CCW’s GGE now has a specific mandate: submit a report to the Seventh Review Conference, scheduled for November 16-20, 2026, with final negotiating sessions in March and August-September 2026. Through 2025, the GGE produced a rolling text with formulations on characterizing LAWS, the applicability of IHL, requirements for human judgment and control, prohibitions on inherently indiscriminate systems, and regulatory measures on predictability and reliability. But whether this text becomes a binding instrument depends on the consensus of states that have spent a decade blocking exactly that outcome.

Meanwhile, the February 2026 REAIM Summit in A Coruna, Spain — the third global summit on responsible AI in the military domain — underscored the divide. Only 35 of 85 attending countries signed the summit declaration. The United States and China both refused to sign, with US Vice President J.D. Vance citing concerns that regulation could stifle innovation and weaken national security.

The Strategic Dimension: Arms Race Dynamics and Proliferation

Beyond the ethical and legal debates, autonomous weapons raise profound strategic stability concerns. An arms race in autonomous weapons is already underway. The US Department of Defense’s Replicator initiative, launched in August 2023, aimed to deploy thousands of attritable autonomous systems by August 2025 — but by that deadline, only hundreds had been fielded, with critical technical and procurement challenges persisting. A second phase, Replicator 2, was announced in September 2024 to focus on countering small unmanned aerial systems. China’s military AI investments continue to accelerate, with PLA testing of advanced drone swarm systems and autonomous ground platforms throughout 2025. The strategic logic is competitive: if adversaries develop autonomous weapons, not developing them creates a perceived military disadvantage.

The proliferation risk is particularly acute for autonomous weapons because the enabling technologies are largely dual-use and commercially available. The AI models, computer vision systems, and drone hardware that power autonomous weapons are derived from civilian technology. A commercial quadcopter, an edge computing processor, and open-source object detection software can theoretically create a rudimentary autonomous weapon. The barrier to entry is far lower than for nuclear, chemical, or biological weapons, making proliferation to non-state actors — terrorist organizations, criminal networks, private military companies — a realistic concern.

The speed of autonomous systems creates escalation risks. Autonomous weapons can operate at machine speed — identifying and engaging targets in milliseconds, far faster than human decision-making cycles. In a contested environment where both sides deploy autonomous systems, the interaction speed may exceed human ability to intervene, understand, or de-escalate. This is the “flash war” scenario: an automated escalation spiral that occurs too quickly for human commanders to arrest. The analogy to algorithmic trading flash crashes in financial markets — where on May 6, 2010, automated systems amplified a market event into a temporary $1 trillion loss in approximately 36 minutes — is instructive and alarming.

Frequently Asked Questions

Sources & Further Reading

• UN CCW Group of Governmental Experts on LAWS
• Campaign to Stop Killer Robots
• ICRC Position on Autonomous Weapons Systems
• UNGA Resolution on Autonomous Weapons (December 2024)
• UN Panel of Experts on Libya – Kargu-2 Report (S/2021/229)
• US DoD Replicator Initiative
• UN Secretary-General Calls for Ban on Killer Robots (May 2025)
• Arms Control Association: Geopolitics and Autonomous Weapons Regulation
• Human Rights Watch – Autonomous Weapons Reports
• ASIL: LAWS and International Law – Momentum Towards a Treaty