AI Is ‘Fuelling the Flames of War’: The Starkest Warnings from the 2025 Shangri-La Dialogue

Humans Removed from the Loop: The Autonomous Strike Reality

One of the dialogue’s most sobering themes was the confirmation that autonomous weapons operating without meaningful human oversight are not theoretical — they are present on today’s battlefields. While the Shangri-La Dialogue itself drove this message home in defense sessions, Singapore’s Foreign Minister, Dr. Vivian Balakrishnan, echoed the same warning at a related UN Security Council debate on AI and international peace. He stated directly that lethal autonomous weapons are “no longer potential,” noting that in hot wars being fought right now, “the truth is because of the urgency to identify multiple targets and to deal with them, human fingers, even today, are often not on the triggers. We have had to outsource it to AI systems” .

The danger, as framed across the dialogue, is twofold. At the tactical level, autonomous systems making life-and-death decisions without human control can produce unpredictable outcomes on fast-moving battlefields. At the strategic level, the removal of human judgment from strike decisions creates the risk of rapid, irreversible escalation — where machines escalate skirmishes into larger conflicts before humans can intervene . This concern threaded directly into the broader nuclear escalation fears that dominated the summit.

South Korea’s ‘Smart Military’ Vision Takes Center Stage

South Korean Defense Minister Ahn Gyu-baek used the summit to position Seoul as a major force in AI-enabled defense, outlining an ambitious “smart military” concept built around artificial intelligence-based systems, drone and anti-drone defenses, and autonomous platforms . The vision goes well beyond slogans, backed by a detailed national effort to embed AI across South Korea’s force structure:

• AI-enabled command and control: South Korea is integrating artificial intelligence across command systems, surveillance, and reconnaissance to accelerate decision-making and reduce reliance on human operators .
• Drones as standard equipment: The South Korean Army envisions drones becoming as common as personal firearms for troops, with a goal of training “500,000 drone warriors” and introducing more than 50,000 training drones by 2029 — enough for one drone per squad .
• Manned-unmanned teaming: At the 77th Armed Forces Day parade, South Korea showcased low-observable unmanned loyal wingman drones, loitering munitions, and robotic combat systems designed for manned-unmanned teaming operations .
• Manpower offset strategy: The AI push is partly a response to declining conscription resources tied to South Korea’s shrinking birth rate. Ahn has stated that AI is “no longer a choice but an essential force element” in restructuring units and personnel .
• Budget commitment: The national AI budget is allocated at 10.1 trillion won, with a broader defense budget of 66.3 trillion won, signaling a major national priority .

South Korea has also taken a diplomatic leadership role, hosting an international summit in Seoul to craft a “blueprint for responsible AI use in the military.” More than 90 countries, including the United States and China, participated. However, any resulting framework is expected to be non-binding, and it remains unclear how many nations will ultimately endorse even a minimum set of guardrails .

The Nuclear Flashpoint: Taiwan

The most alarming assessment to emerge was not directly about AI — but AI’s capacity to accelerate conflict to nuclear levels was at the heart of the IISS’s starkest warning. A dedicated study released during the dialogue concluded that any military conflict between the U.S. and China over Taiwan would risk rapid nuclear escalation. The analysis found that both sides would likely launch sweeping operations targeting each other’s command, control, and communications hubs — creating a powerful destabilizing pressure to escalate to nuclear weapons before losing the capacity to coordinate .

The IISS assessment warned bluntly that a conventional clash over Taiwan could “rapidly snowball into a nuclear crisis” and that the world is “on the cusp of a new nuclear arms race” with the Asia-Pacific at its core . This warning was echoed at the dialogue by U.S. Defense Secretary Pete Hegseth, who warned that a Chinese attack on Taiwan “could be imminent” and reinforced the American deterrence posture . Beijing issued sharp warnings in return, with Chinese representatives reportedly “incensed” by the framing . The nuclear dimension also connects to the AI thread: the integration of autonomous systems with nuclear capabilities means traditional arms-control frameworks are being overtaken without any established control method to replace them .

What the 2025 Shangri-La Dialogue Made Clear

The dialogue painted a picture more urgent than prior summits. The central takeaway was that AI is no longer a future-war concern — it is actively reshaping conflict timelines, removing human judgment from lethal decisions, and eroding the firebreak between conventional and nuclear escalation. South Korea’s “smart military” push shows how middle powers are racing to embed AI before governance frameworks exist, while the Taiwan nuclear-risk assessment underscores that the region’s primary flashpoint carries escalation dynamics that no existing arms-control structure is designed to handle.

For the defense community and the public, the dialogue left a clear message: AI in warfare is outpacing both governance and strategic doctrine, and the time to establish boundaries has already slipped into the rearview mirror.