AI Raises Risks in Modern Warfare

Artificial intelligence is rapidly transforming modern warfare, but military officials and security experts are increasingly concerned that unchecked AI development could make global conflicts faster, less predictable and more dangerous.

At the Beijing Xiangshan Forum, China’s flagship defence conference, international delegates highlighted risks associated with autonomous weapons, AI-driven misinformation and the growing integration of intelligent systems into military decision-making.

One major concern is compressed decision time. AI can analyse battlefield information and recommend responses almost instantly, potentially leaving commanders and political leaders with less time to verify intelligence, assess consequences or prevent an unintended escalation.

The risks become particularly serious when AI-generated deepfakes, synthetic media and misinformation enter military environments. Manipulated audio, video or intelligence could create confusion over an adversary’s intentions or even generate false signals during a geopolitical crisis.

Autonomous weapons present another challenge. As machines gain greater ability to identify targets, navigate independently and execute complex missions, delegates stressed the importance of maintaining meaningful human oversight, especially over decisions involving lethal force and strategic weapons.

The debate has added significance as the United States and China, the world’s leading AI powers, prepare for bilateral talks while remaining divided over how military AI should be governed. Nuclear command-and-control systems represent an especially sensitive area where uncontrolled automation could carry enormous consequences.

The emerging challenge is therefore bigger than technological superiority. Military AI must combine capability with accountability, cybersecurity, authenticity and human control. In an era of autonomous machines and synthetic information, the critical question is not simply how powerful AI becomes—but who remains responsible for the final decision.