TL;DR
At the Shangri-La Dialogue in Singapore, senior military officials warned that AI is compressing battlefield decision-making faster than humans can process, eclipsing nuclear weapons as the dominant strategic concern. Ukraine and the US-Iran conflict were cited as live examples of AI already shaping combat operations.
The dangers of artificial intelligence eclipsed nuclear weapons as the central concern at a strategic stability panel during the Shangri-La Dialogue in Singapore, with senior military officials warning that AI-driven systems are collapsing the time available for human decision-making in conflict. The annual defense summit, held from 29 to 31 May, drew defence ministers and military chiefs from across the Indo-Pacific and beyond.
Lieutenant General Nauman Zakria, Commander of 1 Corps and the Army Rocket Force Command of the Pakistan Army, framed the threat in terms of the OODA loop, the military decision-making cycle of observe, orient, decide, and act. AI compresses that loop to the point where it creates a fog in which “a human can’t evaluate the situation fast enough,” he said. “People will act irrationally, and the actions will be extreme.”
Already on the battlefield
The warnings were not theoretical. General Onno Eichelsheim, the Netherlands’ chief of defence, noted that AI had already appeared in active combat. Ukrainian forces have deployed AI systems to anticipate Russian attacks and coordinate drone operations across the front line, using machine learning trained on years of battlefield footage to identify targets and adapt to countermeasures in real time.
The United States has also acknowledged using AI tools in planning strikes against Iranian targets. The Pentagon confirmed that warfighters leveraged “advanced AI tools” to sift through data and make faster targeting decisions during Operation Epic Fury, which has struck more than 13,000 targets since it began.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!
“AI is a huge risk in escalation. I think that’s clear,” Eichelsheim said. “But I’m not naive. It’ll be used in the domain. It is already being used.”
The ICRC warning
Mirjana Spoljaric, president of the International Committee of the Red Cross and the only panellist not directly involved in defence, offered the sharpest assessment of AI’s humanitarian risk. She warned that while technology had the capacity to improve lives, it was vastly increasing the danger of war in practice.
“We don’t know where the trigger is pulled,” Spoljaric said. “It could be thousands of kilometres away. So while there are potentials of AI for protecting civilians, what we see at the moment is only the negative side.”
The ICRC has long argued that autonomous weapons systems must retain meaningful human control over targeting decisions, a position that is gaining urgency as AI-enabled systems move from experimental programmes into active deployment. No major military power has committed to binding restrictions on autonomous weapons, despite years of discussions at the United Nations Convention on Certain Conventional Weapons.
China’s nuclear and AI positions
Nuclear weapons still featured in the conversation. People’s Liberation Army Major General Meng Xiangqing reaffirmed China’s no-first-use policy and proposed that all five recognised nuclear-weapon states negotiate a mutual no-first-use treaty. “If we can do so, we can reduce the risk and we can further enhance strategic stability,” he said.
China has also called for international rules governing the military use of AI, including eventually binding legal frameworks. But its own position on autonomous weapons remains ambiguous. Beijing published a position paper on military AI regulation that makes no mention of restricting lethal autonomous weapons systems, and the PLA is investing heavily in AI-enabled military capabilities.
The compression problem
The thread running through the panel was not that AI itself is inherently dangerous, but that its speed creates a structural problem for conflict management. Traditional deterrence assumes that decision-makers have time to assess information, consult allies, and weigh consequences before acting. AI-enabled systems can identify targets, recommend responses, and execute operations faster than any human oversight process can function.
That compression matters most at moments of escalation. If one side’s AI systems detect an incoming threat and recommend an immediate response, the decision-maker may have seconds rather than minutes or hours to evaluate whether the threat assessment is accurate. Military analysts have warned for years that this dynamic could turn miscalculations into full-scale conflicts before anyone has time to intervene.
The Shangri-La Dialogue did not produce any new agreements or binding commitments on military AI, and efforts to establish legal frameworks for AI governance remain fragmented across jurisdictions. But the fact that a panel ostensibly about strategic stability, traditionally a discussion about nuclear posture and missile defence, spent most of its time on artificial intelligence reflects a shift in what the world’s defence establishments now consider the most pressing threat to international security. The question is no longer whether AI will be used in warfare. It is whether humans will retain enough control over it to prevent catastrophic mistakes.
