The United Nations will launch on Thursday its first international forum dedicated to the military uses of artificial intelligence, an initiative of the UN Institute for Disarmament Research that has drawn funding from all five major nuclear powers: the United States, Russia, China, Britain and France. Robin Geiss, who heads the Geneva-based institute known as UNIDIR, told a press conference on Tuesday that "the margin for human judgment is increasingly narrow, including in the nuclear domain." The forum is meant to be the first standing venue where governments, companies and researchers address military AI together.
Geiss was candid that he could not say how close the world is to machines acting outside human control. AI is "creeping into so many areas," he said, from early-warning systems to reconnaissance, that "it's quite difficult to say at which point we're actually really losing control." He framed the launch as a turn from years of reluctance to set guardrails toward "a growing realization that we stand at a crossroad."
The forum is one of several efforts the institute's new Center of Excellence on AI, Peace and Security is rolling out this week. A second will convene defense-technology companies, software developers, subcontractors and investors to draft "practical, voluntary guidelines" for responsible industry behavior, on the argument that corporate design choices determine whether military AI is "reliable, predictable and whether they can operate in line with international law." A third will address the misuse of AI by armed groups, which Geiss warned are already using it for operational planning and logistics.
The common thread is that the hardest problems are being handed to voluntary, consultative bodies rather than binding rules. Geiss acknowledged that the global race to lead on AI is a serious obstacle, but argued that "concerns about AI's impact are already transcending geopolitical blocs." He pointed to a proposal last month for the United States and China to set up an AI safety alert system for issues that could threaten national security.
That the five nuclear-weapons states all helped fund the center is the most concrete sign of shared concern, and also its clearest limit. The same governments competing hardest to field military AI are the ones underwriting the forum meant to restrain it, and nothing announced on Thursday carries the force of a treaty. Voluntary frameworks have a long history in arms control, sometimes as a first step toward binding limits, sometimes as a substitute for them.
What the initiative will actually do remains vague, as Geiss conceded. The test will be whether a standing forum can produce shared definitions and redlines faster than the technology erodes the "margin for human judgment" he described, and whether industry guidelines with no enforcement can shape companies racing the same clock as the states that buy from them. For now the UN has done the achievable thing and created the room. Filling it is the harder task.