
The Day Diplomacy Blinked: How the World Is Sleepwalking into an AI Arms Race
While diplomats in Geneva negotiated yet another timeline to discuss lethal autonomous weapons, AI-enabled drones were already identifying, tracking and striking targets on contemporary battlefields. The future the United Nations hopes to regulate by 2028 has already begun. That is the defining contradiction of our age: diplomacy measures time in conferences and consensus, while artificial intelligence evolves through software updates. The latest agreement under the UN Convention on Certain Conventional Weapons (CCW), committing states to merely "elaborate" rules by 2028, is therefore less a diplomatic breakthrough than an admission that global governance is falling behind technological reality.
The deeper problem is structural. For thirteen years, negotiations under the CCW have been trapped by a consensus rule that allows a handful of military powers to delay collective action. More than a hundred countries sought negotiations on a legally binding treaty within a year, but procedural resistance prevailed. This is not accidental. None of the major military powers wants to accept constraints while rivals continue investing in military AI. Every state fears that regulation today may become strategic disadvantage tomorrow. The result is a classic security dilemma: nations accelerate the very technologies they simultaneously warn could threaten humanity.
Artificial intelligence represents more than another military innovation. Like gunpowder, industrial warfare and nuclear weapons before it, AI is reshaping the character of war. Yet unlike those revolutions, this one spreads through software rather than factories. AI-enabled targeting systems, autonomous drone swarms and machine-assisted battlefield decisions are no longer futuristic concepts but operational realities. Recent military demonstrations of robotic combat vehicles coordinating autonomous drone attacks illustrate how rapidly decision-making is shifting from soldiers to algorithms. Diplomacy is negotiating tomorrow's treaty while technology is rewriting today's battlefield.
The economics of this transformation make regulation even harder. Governments are no longer the sole architects of warfare. Defence contractors, AI start-ups, semiconductor firms and venture capital increasingly shape military innovation. Once billions are invested in autonomous systems, political incentives shift from restraint to commercial protection. An emerging military AI economy now has a direct stake in ensuring regulation never outpaces innovation.
Nor can autonomous weapons be controlled like nuclear arsenals. Uranium can be inspected. Missile silos can be counted. Software cannot. Open-source AI models, commercial computing power and globally available chips have dramatically lowered barriers to military innovation. Verification, the cornerstone of successful arms-control agreements, becomes extraordinarily difficult when the weapon exists primarily as code. Treaties designed for industrial-age hardware are poorly suited to digital-age algorithms.
The strategic consequences are equally profound. Future conflicts will not separate autonomous weapons from cyberattacks, electronic warfare and space-based intelligence. AI is becoming the operating system connecting them all. As machines increasingly respond to machines, the time available for political judgment shrinks dramatically. Nuclear deterrence survived partly because leaders retained precious minutes to assess escalation. Autonomous systems threaten to eliminate that pause. Algorithms reacting to algorithms could trigger conflicts that commanders neither intended nor fully understand until missiles are already in flight.
This dynamic also weakens deterrence itself. If military planners conclude that an adversary's autonomous systems can disable command networks within minutes, the temptation shifts towards pre-emptive action rather than restraint. At the same time, autonomous systems reduce the political cost of using force by limiting casualties among one's own soldiers. A technology designed to make war more efficient may ultimately make it more frequent.
The greatest victims, however, may not be the powers driving this race. Countries with the least influence over military AI development, particularly fragile states across the Global South, are likely to become testing grounds for autonomous warfare. As drones become cheaper and algorithms more accessible, the technology will inevitably diffuse beyond state militaries to criminal networks and terrorist organisations. History suggests that military innovation rarely remains exclusive for long.
Beyond strategy lies a more fundamental question of civilisation. Every society has treated the decision to take human life as requiring moral agency. Delegating that responsibility to statistical optimisation transforms killing from a grave human burden into an engineering problem. International humanitarian law demands judgment, proportionality and context, qualities machines cannot genuinely possess. AI excels at recognising patterns but struggles with ambiguity, yet battlefields are defined by ambiguity. Civilians flee unpredictably, combatants surrender unexpectedly and environments change in seconds. Statistical confidence should never become a substitute for moral certainty. Nor does the law answer the inevitable question of accountability: when an autonomous system commits a war crime, who stands trial?
None of this argues against artificial intelligence itself. AI can strengthen logistics, surveillance, disaster response, mine clearance and force protection. The line the world must defend is simpler: life-and-death decisions should never be delegated to machines without meaningful human control.
The tragedy is that the institutions charged with preventing this future are themselves trapped in the past. The CCW was designed for an era when weapons evolved over decades. Artificial intelligence evolves every few months. Institutions built for industrial-age diplomacy cannot regulate technologies developing at digital speed. The world should therefore pursue an Ottawa-style coalition of willing states to establish binding international norms, rather than allowing consensus to become an excuse for paralysis.
History rarely announces technological turning points in advance. They become obvious only after norms have collapsed and institutions realise they were negotiating yesterday's problem. If world leaders continue treating 2028 as a destination rather than a deadline, they will not be negotiating the rules of autonomous warfare. They will be writing the obituary of meaningful human control over war.
