In the Cold War, arms control meant counting missiles and warheads.

How does that model work in the cyber century, the drone decades and the AI age? What do you count, and how do you control it?

The old control mechanisms created last century have crumbled, while the new age asks ever-harder questions.

The annual Shangri-La defence summit, convened in Singapore by the International Institute for Strategic Studies from 30 May to 1 June, heard repeated calls for returning to arms control. But it’s an aspiration without a firm plan for achievement.

In a session on managing proliferation risks, Australian Defence Minister Richard Marles said the Indo-Pacific had the world’s largest conventional military and nuclear rearmament, and was ‘ground zero in the race for technological supremacy’.

Yet Marles observed that the global architecture that ‘provided a foundation for strategic deterrence and assurance has fallen into dangerous decline’.  The Intermediate-Range Nuclear Forces Treaty collapsed in 2019, leaving non-strategic nuclear weapons unchecked by a formal agreement, he noted. In 2023, Russia suspended its participation in the last remaining binding bilateral arms control treaty between itself and the United States, NewSTART, with the treaty itself set to expire in February 2026. This leaves no legally binding limits on the strategic nuclear arsenals of the two largest nuclear powers for the first time since 1972.

Marles said arms control today was vastly more complicated and the risks were growing:

The arms control regime that we knew was a foundational component of an oft-maligned concept—the international rules-based order.  Arms control must be seen as a necessary but not sufficient feature of a broader strategic order that we must build anew, an order defined as much by rules and norms as it is by power.

Japanese Defence Minister Gen Nakatani put the control problem in the broadest terms, worrying about what he called an abdication of responsibility for the international public good:

States that do not share our values and interests are prepared to undermine the very foundations of our societies and economies for their own gain. State-sponsored cyber-attacks are on the rise, and repeated damage to undersea cables, suspected to involve certain countries, these have had serious implications in the field relevant to defence.

Dutch Defence Minister Ruben Brekelmans said his nation had organised the 2023 Responsible AI in the Military Domain summit calling for robust international agreements because ambition for AI should not outpace responsibility. The military challenge was to innovate and to regulate, to lead and to safeguard, Brekelmans said:

What happens when AI misreads intent? When escalation ladders are climbed not by men, but by machines? When command systems are hacked, spoofed or deceived? These are not hypothetical risks. These are very real. And they are compounded by cyber threats, misinformation, anti-satellite weapons, all tools that are already being used to destabilise, deceive and divide.

French Armed Forces Minister Sebastien Lecornu said the challenge in the space, cyber, underwater and information domains was that the environments could be both civilian and military. He said the aim must be to regulate and control conflict domains: ‘If I had to summarise our dilemma today, you can be defeated without being invaded and you can lose the war without ever having been at war. That is really a brutal new strategic deal compared with the situation we used to know.’

As the Shangri-La dialogue was posing questions as much as reaching for answers, one of today’s great strategists, Sir Lawrence Freedman, was posting an article asking: ‘whatever happened to arms control?’ Freedman offered three problems challenging control efforts:

—The difficulties in getting agreement on force levels and types of weapons.

—Instead of the bilateral Cold War negotiations between Moscow and Washington, agreements today would, at the least, also involve Beijing, London and Paris, and ‘anything of universal application has to take account of up to 200 states’.

—Weapons are becoming harder to count (for example, drones) while the new capabilities of cyberattacks and AI ‘are as civilian as they are military’.

Freedman’s conclusion is that achieving arms control needs a basis of trust:

When two sides habitually see the worst in each other, it is never going to be easy to get them to agree to provisions which require assuming the best. If trust is needed then that requires resolving political differences and working on improving relations, and sometimes the conflicts run too deep for that to be possible. That is where we are now. Peace does not come through arms control: arms control comes through peace.