
A Two-Year Path for Rules on Machine-Led Force
Governments have agreed to continue negotiations aimed at developing new international rules for lethal autonomous weapons by 2028. The decision keeps diplomacy alive after years of debate over military systems that can identify, select or attack targets with limited human involvement. It also exposes the central divide in the process: many countries want a legally binding instrument, while several major military powers remain unwilling to accept that commitment.
The outcome is therefore both a procedural advance and a warning. States did not abandon the issue, and they created a timetable for further work. Yet they stopped short of promising a treaty or defining the level of human control that must remain over decisions to use lethal force. As artificial intelligence is integrated into surveillance, targeting and weapons platforms, the gap between technological capability and international regulation is becoming harder to ignore.
What Countries Agreed and What They Left Open
The agreement calls for further discussions to elaborate possible rules during the next two years. It follows negotiations among states belonging to the Convention on Certain Conventional Weapons, a framework designed to restrict weapons that may cause excessive injury or indiscriminate effects. Around one hundred countries supported stronger and legally enforceable limits, but opposition from major arms-producing states weakened the final language.
The process has not decided whether the eventual result will be binding, voluntary or a mixture of prohibitions and best-practice standards. That ambiguity allowed consensus, but it also means governments can interpret the timetable differently. Supporters of a treaty see 2028 as a deadline for urgent negotiation. Skeptical states may treat it as an invitation to continue technical discussion without accepting new legal duties.
Why Human Control Is the Core Question
Autonomous systems cover a wide range of technologies. Some assist human operators by detecting objects or recommending possible targets. Others may navigate, track or engage with less direct supervision. The most serious ethical and legal concern arises when a machine's assessment can lead to death without a person making a meaningful, informed and timely decision about the individual target and circumstances.
International humanitarian law already requires distinction between civilians and combatants, proportionality in attack and precautions to reduce civilian harm. The unresolved question is whether an autonomous system can apply those duties reliably in complex, changing environments. Accountability is equally difficult. If an algorithm makes an unlawful targeting decision, responsibility may be dispersed among commanders, programmers, manufacturers and political leaders. A rule that merely requires a human to approve a machine recommendation may be inadequate if the operator lacks time, information or a realistic ability to refuse.
Military Competition Is Moving Faster Than Diplomacy
Armed forces are investing in drones, sensor networks, automated defence systems and software that can coordinate operations at high speed. Recent conflicts have demonstrated how inexpensive unmanned systems can change battlefield tactics, while AI can process imagery and other data faster than human teams. These developments create strong incentives for states to deploy increasingly independent weapons before common safeguards exist.
Military planners argue that automation can improve precision, protect personnel and respond to threats too quickly for manual control. Critics answer that speed can also magnify errors, trigger escalation and make lethal decisions less transparent. Systems trained on incomplete data may misclassify objects or people, and adversaries may deceive sensors. The risk is not limited to a single malfunction. Large numbers of networked systems could produce cascading decisions that humans struggle to understand or stop.
The Choices That Will Define the 2028 Outcome
Future negotiations will need to move beyond general principles. Governments must decide which autonomous functions should be prohibited, what level of human judgment is mandatory, how systems must be tested and reviewed, and how violations will be investigated. They will also have to address weapons that learn or change after deployment, as well as software upgrades that can alter a system's behaviour without changing its physical platform.
The latest agreement proves that deeply divided states can still preserve a negotiating channel. It does not guarantee a strong result. Civil-society groups and humanitarian organisations fear that a two-year process could become a delay while deployment accelerates. The credibility of the talks will depend on measurable progress, clear definitions and a willingness to place legal boundaries around military advantage. By 2028, the world may possess far more capable autonomous systems than it does today. The rules must be ready before those machines, not after them.
Verification will be another difficult part of any future regime. Unlike a missile or aircraft, decisive software can be copied, updated or hidden inside a broader platform. Effective rules may therefore require national reviews, testing records, audit access and reporting when systems fail. They must also protect legitimate security information without turning secrecy into an escape from accountability. A framework that cannot be checked in practice may sound reassuring while doing little to restrain deployment. Civilian protection must remain the measure of success.
PUBLISHED BY SUYASH PACHAURI, FOUNDER & OWNER, GLOBAL BOLLYWOOD | THE HOLLYWOOD SCOPE.