Meher khan
Three years ago, UN Secretary-General António Guterres set a deadline: by 2026, the world would have a legally binding treaty banning autonomous weapons that kill without human oversight. That deadline has now passed. There is no treaty. There is not even a negotiating table, only the same consultations that have circled Geneva for over a decade.
This should alarm anyone who thinks international law still matters in an age of algorithmic warfare. Autonomous weapons systems select and engage targets using sensor data rather than a human finger on a trigger. They are no longer speculative. They are being tested, deployed, and refined in active conflicts, while the treaty meant to govern them remains stuck in permanent draft.
The paralysis is structural, not accidental. Talks have run since 2014 inside the Convention on Certain Conventional Weapons, a forum that operates by consensus, meaning any single state can block progress indefinitely.
The United States has opposed a binding instrument outright, and Japan has resisted a legally binding track. Meanwhile, support for regulation is not fringe: more than a hundred countries now back new rules, 166 states voted in December 2024 to keep the issue on the UN General Assembly’s agenda, and in September 2025, forty-two states jointly pushed for negotiations to begin on the basis of existing draft text. The problem is not a lack of consensus among the many. It is a veto held by the few.
This is where the case for abandoning the consensus-based process gets serious. International law has faced this exact bottleneck before, and broken it before. When consensus forums failed to ban landmines and cluster munitions, coalitions of willing states moved negotiations elsewhere and produced the Ottawa and Oslo conventions without waiting for holdout powers.
The Treaty on the Prohibition of Nuclear Weapons followed the same logic. Autonomous weapons need the same move: a majority-backed, General Assembly-anchored negotiating track that does not grant any one military power a permanent veto over whether machines are allowed to decide who dies.
Critics of this approach warn that a treaty without the major developers of these systems on board is symbolic at best. That concern is not unreasonable, but it mistakes universality for effectiveness. A norm against a weapons category does not need every state’s signature to reshape behaviour, it needs enough weight to stigmatise the technology, restrict its proliferation, and give courts, procurement officers, and export-control regimes a legal standard to enforce against. Landmines were not eliminated because every state signed Ottawa; they were marginalised because using them became internationally costly even for non-signatories.
There is also a Global South dimension to this debate that keeps being treated as an afterthought. Autonomous weapons are disproportionately likely to be tested and used in conflicts fought on the territory of, or by proxy through, states outside the major military powers Libya’s 2020 conflict, where a Turkish-made drone reportedly engaged retreating forces with no confirmed human command link, remains the clearest warning so far.
Pacific and developing states have explicitly asked the UN for an immediate start to binding negotiations, citing risks of misuse by non-state actors and algorithmic bias compounding existing inequalities. These are not hypothetical concerns raised by distant academic panels; they are the direct submissions of states that expect to inherit the consequences of a governance vacuum they did not create.
None of this requires perfect global unanimity to matter. It requires states that already agree, a group that on paper is a clear majority, to stop waiting for permission from states that have made clear they will never give it. The consensus rule was designed to protect deliberation, not to hand a small number of governments a permanent veto over whether the world regulates machines that kill. Nine years of talks without result is not deliberation. It is obstruction with better manners.
The 2026 deadline was never going to enforce itself. But missing it should not be read as proof that autonomous weapons governance is impossible, only that it will not happen inside a room where one state’s objection outweighs a hundred states’ consent. The next move belongs to the General Assembly, not the forum that has spent nine years failing to make one.
Meher Khan is a Junior Paralegal in the Industrial Diseases Pre-Litigation team at Baker & Coleman Solicitors and holds an LLB (Hons) from the University of London. She is the founder of Legally.Loud, a digital platform on technology law and digital rights in Pakistan, and writes on AI governance, cyber policy, and international law from a Global South perspective.
