The United Nations Institute for Disarmament Research plans to launch an international forum on military uses of artificial intelligence, bringing together governments, technology companies and other participants to address risks ranging from diminished human control to misuse by armed groups.
The forum, scheduled to launch Thursday, will sit within UNIDIR’s Center of Excellence on AI, Peace and Security. It is part of a broader set of initiatives that includes voluntary guidance for industry and work on the risks posed by armed groups using AI.
At a U.N. press conference Tuesday, UNIDIR director Robin Geiss warned that the growing integration of AI into military systems could leave less room for human judgment, including in matters involving nuclear weapons. He said uncertainty about when these systems might move beyond effective human control is itself a central problem the initiative seeks to address.
The concern extends beyond autonomous weapons. AI is being incorporated into early warning, intelligence collection and reconnaissance, making it difficult to identify the point at which automated assistance begins to erode meaningful human oversight. Those functions can influence military decisions even when a person retains formal authority to act.
Technically, keeping a human involved in a decision does not by itself resolve every control problem. Operators must also be able to understand a system’s limitations, evaluate its recommendations and intervene when necessary. In time-sensitive settings, the speed and volume of machine-generated information can complicate those tasks. These are broader challenges for military AI, rather than capabilities or standards announced for the new forum.
Geiss acknowledged that competition among countries to lead in AI creates obstacles to international cooperation. But he pointed to signs of shared concern across geopolitical divisions, including a proposal for the United States and China to establish an AI safety alert system for problems that could threaten national security.
The Geneva-based institute established its Center of Excellence on AI, Peace and Security this summer as the U.N.’s first permanent platform dedicated to those issues. Geiss said the center received funding from all five major nuclear powers: the United States, Russia, China, Britain and France.
Details of the forum’s work remain limited. Geiss described a body drawing participants from relevant sectors around the world, but did not set out a detailed program or specific expected outcomes.
A parallel industry initiative will seek to develop practical, voluntary guidelines for responsible conduct consistent with international law and human rights. Its focus reflects the influence that commercial design and development decisions can have on whether military AI systems behave reliably and predictably.
UNIDIR plans consultations this week involving defense technology companies, software developers, subcontractors and investors, alongside governments, civil society, academics and other experts. The framework is intended to address industry behavior; it was not presented as a binding international agreement.
The institute is also launching work on AI misuse by armed groups. Geiss said such groups are already using the technology to assist operational planning, execution and logistics, placing those applications alongside state military systems on the center’s agenda.
Sources: abcnews.com