The UN Institute for Disarmament Research launched a forum on October 22 that gives governments, defence companies, investors, academics and civil society one room to argue about military AI.
It sits inside a Centre of Excellence on AI, Peace and Security that UNIDIR established this summer. Its director, Robin Geiss, said the centre is funded by all five major nuclear powers, which is the detail that makes the venue unusual rather than merely formal.
Geiss framed the concern in terms of distance. The margin for human judgment is narrowing, he said, and that includes the nuclear domain. He pointed at early-warning systems, intelligence gathering and reconnaissance, and said it has become hard to say at which point control is actually lost.
The forum has no power to bind anyone. What it can do is build a shared vocabulary for identifying risk and comparing national policy, and it is drafting voluntary guidance for companies whose design decisions determine whether a system is predictable and consistent with international law.
That industry track carries a quiet tension. Defence contractors and their subcontractors are being asked for input into rules they would later be measured against.
Earlier UN discussions settled one principle, that decisions on nuclear use stay with humans. Extending that line into targeting and reconnaissance is the harder negotiation, and it is the one this forum now hosts.
