Geneva's first Global Dialogue on AI Governance opened with a call to ban machines that pick targets without a human. The fight now is whether the rules come from diplomats, courts, or militaries.
António Guterres stood before diplomats in Geneva on Monday and called, in the plainest language the United Nations owns, for an international ban on "killer robots." He described machines that "select and engage targets and take a life without human control and judgment" as "morally repugnant" and said choices about life and death "must remain forever human" (UN Secretary-General remarks).
Guterres opened the first Global Dialogue on AI Governance, convened by the UN in Geneva, with more than 4,000 delegates from governments, industry, and civil society in the room. The Dialogue is the institution's first attempt to convert more than a decade of "killer robots" rhetoric into a binding international-law instrument, or, failing that, a political declaration that could outlive the meeting.
The fight is over who writes the rules for autonomous war. The two leading military powers, the United States and Russia, have already blocked two years of UN General Assembly resolutions on military AI. Their position, broadly, is that existing international humanitarian law is enough. Guterres rejects that position explicitly: speed, not legality, is the problem.
"An experiment is being run on our own societies, without a plan and without consent," Guterres said in his opening remarks. "That is not sustainable." The phrase captures the speed mismatch the Dialogue is supposed to address. Militaries are folding AI into targeting, intelligence, and command-and-control at a pace that no multilateral process has matched. The weapons the Secretary-General wants banned are not future systems in a lab; they are decisions about engagement being pushed downward, away from human commanders and toward algorithmic pipelines.
As reported in Newser's summary of WSJ reporting, the AI lab Anthropic is locked in a court fight with the Pentagon after seeking contractual guarantees that its models would not be used for autonomous targeting or domestic surveillance. The Pentagon's response, that it needs access for "any lawful use," is the position Guterres is now trying to outlaw in Geneva. The Anthropic case is the structural proof: when international law stalls, the norm-setting fight moves into procurement contracts and federal courtrooms.
The moral coalition is widening in parallel. Pope Leo XIV has argued, according to UN background material, that AI weapons risk making war easier to wage and harder to end, casting the question as one of human control rather than technological inevitability. The civil-society Campaign to Stop Killer Robots has been pressing the same argument for more than a decade, with limited success at the negotiating table.
The Dialogue runs through Wednesday. A political declaration is achievable; a treaty is not. The realistic outcomes are a shared set of principles that future negotiations can build on, or a communiqué that codifies disagreement and cedes the field.
Guterres's speech on Monday was not a request for consensus. It was an attempt to put the burden of proof on the governments integrating AI into their kill chains, and to do so before the integration is complete enough to be reversed.