UNIDIR director Robin Geiss warned in New York on October 20 that AI is increasingly displacing human decision-making in warfare. According to the official UN account, his concern extends to the nuclear domain. At the press briefing, he asked whether machines should be allowed to select and kill human targets.
Geiss described shorter intervals between identifying and striking targets, alongside opaque systems, as escalation risks. Associated Press also reported his warning that it is difficult to establish the point at which human control is actually lost. He was describing an uncertain trajectory, not presenting a newly investigated individual attack.
He announced a new forum for October 22 through UNIDIR’s Centre of Excellence on AI, Peace and Security. Its planned work includes voluntary guidelines for responsible industry behaviour, a shared AI-security vocabulary and addressing potential misuse by non-State armed groups.
Those activities were still ahead of the forum when he spoke. Geiss also identified geopolitical obstacles to common rules. Whether the announced discussions produce concrete, effective safeguards remains open.