The time between identifying a military target and striking it is shrinking as artificial intelligence takes on more battlefield tasks, Robin Geiss, director of the UN Institute for Disarmament Research, warned Tuesday.
“Are we going to let machines target and kill human beings?” Geiss asked during a briefing at UN Headquarters in New York, warning that AI was increasingly “on track to crowding out human decision-making” in warfare, including the nuclear domain.
Geiss said militaries are using AI to analyze intelligence, identify targets, operate drones, and coordinate cyberattacks. Many of these systems are difficult to interpret, making errors harder to spot and increasing the risk of miscalculation and unintended escalation, he added.
Last month, an AI chatbot reportedly helped produce an intelligence report wrongly identifying a Chinese vessel's cargo as nuclear materials. U.S. forces prepared to board the ship and aircraft scrambled before the error was discovered and the operation was called off. One source told CNN the incident “almost started a war.”
UN forum to address military AI risks
UNIDIR plans to launch a new forum on Thursday focused on AI in international peace and security. Geiss said the initiative will develop voluntary guidelines for military AI tools development and clarify terminology around security. It will also examine how militia groups might use the technology for planning and logistics.
Geiss acknowledged geopolitical competition as a hindrance to regulation, but said concerns were being shared across political divides, including among major powers.
In September, the U.S. proposed an AI notification mechanism with China to warn each other about incidents that could threaten national security. The UN Secretary-General has also called for stronger regulation of military AI and a legally binding treaty on lethal autonomous weapons.