‘Whoever wins AI, wins’: How artificial intelligence is transforming the future of war

It has become one of the defining slogans of the current arms race: whoever wins AI, wins. The phrase, repeated by politicians, generals and defence executives alike, compresses a sprawling technological competition into four words — and captures a belief now shaping military budgets, doctrine and diplomacy across the world.

The claim is deliberately sweeping. But the logic behind it is straightforward. Modern warfare generates more information than any human staff can process: satellite imagery, drone video, radar tracks, signals intercepts, open-source social media. The side that can sift that flood fastest, and turn it into decisions, gains an advantage that no amount of steel or manpower can easily offset. Artificial intelligence promises exactly that — speed.

From the battlefield to the back office

Much of the transformation is less cinematic than the phrase “killer robots” suggests. AI is being folded into logistics, predictive maintenance for aircraft and vehicles, translation and intelligence triage, cyber defence, and the planning tools that help commanders weigh options. These are unglamorous functions, but they are where the earliest and most reliable gains appear.

The sharper end is harder to ignore. Uncrewed aerial and naval systems have moved from niche capability to central feature of contemporary conflict, and the software steering them is growing more autonomous — able to navigate, identify targets and operate when communications links are jammed. Swarming tactics, in which dozens or hundreds of cheap drones coordinate with minimal human input, have shifted from research papers into operational planning.

That shift raises the question that dominates every serious debate about military AI: how much judgement should be delegated to a machine, and at what point in the chain does a human remain meaningfully “in the loop”?

The accountability problem

International humanitarian law requires combatants to distinguish between civilians and fighters, and to weigh proportionality. Those are contextual, deeply human judgements. Critics argue that systems trained on incomplete or biased data may fail in precisely the ambiguous situations where the law matters most — and that when they fail, responsibility becomes diffuse. Was it the operator, the commander, the procurement officer, the engineer, or the training dataset?

Campaign groups have pressed for binding international rules on autonomous weapons for years, with limited success. The pace of commercial AI development has outstripped diplomacy, and states that believe they are ahead have little incentive to restrict themselves.

An arms race with private-sector engines

Unlike the nuclear era, this competition is not driven mainly by state laboratories. The foundational advances come from commercial technology firms, and defence ministries increasingly find themselves as customers rather than originators. That dependence has blurred old boundaries between Silicon Valley and the military, and has prompted internal dissent among engineers uneasy about their work’s applications.

It also means the technology proliferates quickly. Capabilities that once required a superpower’s resources can be assembled by smaller states and, in cruder forms, by non-state groups.

The risk inside the slogan

The danger of “whoever wins AI, wins” is that it frames caution as surrender. Compressed decision timelines, opaque systems and mutual suspicion are a recipe for miscalculation — machines escalating faster than diplomats can intervene.

The technology is not going to be uninvented. The open question is whether governments can build guardrails at anything like the speed they are building capabilities. Read More