UN disarmament institute chief warns AI is crowding out human control in warfare

UN News covers remarks to reporters in New York by Robin Geiss, Director of the UN Institute for Disarmament Research (UNIDIR), who warned that artificial intelligence is on course to push out human decision-making in warfare. He asked journalists whether the world will let machines target and kill human beings. The report adds that the Secretary-General has repeatedly urged that this moral red line not be crossed.

Geiss described AI taking on a widening set of military tasks, from analysing intelligence and identifying targets to operating drones and coordinating cyber-attacks, at the cost of human judgement, including in the nuclear field. He said this speeds up operations, shortens the gap between spotting and striking a target, and lets forces engage far more targets each day. He said opaque systems raise the danger of error, miscalculation and unintended escalation, and that easier tools will help non-State actors run sophisticated operations.

Geiss acknowledged that geopolitical rivalry is a serious obstacle to regulation, but said worries about the risks are shared across political outlooks, including among major powers. As an example, he noted that in September Washington suggested to Beijing an AI notification mechanism, through which the two biggest military spenders would alert one another to AI incidents posing a national security threat.

The report says a UNIDIR centre focused on AI, peace and security will open a multi-stakeholder forum on Thursday, described as the first of its kind in the UN system. It will work on voluntary industry guidelines for military AI design, an AI security lexicon to clarify terms such as autonomy, and the risk of misuse by non-State armed groups.