--- title: "Will AI Control Nuclear Weapons? | The Sentient World" description: "Nuclear powers and more than 60 countries have pledged to keep a human in control of nuclear weapons. The UN has voted on it, but nothing binds every state." url: "https://thesentient.world/blog/ai-and-nuclear-weapons" --- # Will AI control nuclear weapons? Nothing in the public record we have read says any government plans to let AI decide to use nuclear weapons. Several have said the opposite. The US and Chinese presidents affirmed in November 2024 that humans must keep control of that decision, and more than 60 countries backed a similar line in 2024. These are pledges and UN votes, and none of the sources we read describes a treaty that binds every nuclear-armed state. October 13, 2026 5 min read By The Sentient World ![The question "Will AI control nuclear weapons?" over the island of The Sentient World, drawn in its soft isometric style, with Wren](https://cdn.thesentient.world/blog/img/auto/ai-and-nuclear-weapons-hero.bca67cba50.jpg) The island of The Sentient World on the day we wrote this. ## Who decides to use nuclear weapons today? In the United States the decision to start or end the use of nuclear weapons belongs to the President. The 2022 US Nuclear Posture Review, the document that sets out the country's nuclear policy, committed to keeping a human in the loop for every action critical to those decisions. That covers the information that reaches the President as well as the carrying out of an order. > “maintain a human 'in the loop' for all actions critical to informing and executing decisions by the President to initiate and terminate nuclear weapon employment” > > 2022 US Nuclear Posture Review The United States is not alone in putting this in writing. In a joint working paper for the 2022 review of the nuclear Non-Proliferation Treaty, France, the United Kingdom and the United States pledged to keep human control and involvement in all actions critical to decisions on using nuclear weapons. The sources we read do not describe how every nuclear-armed country runs its own chain of command, so we cannot say how each of them decides. ## Have the US and China agreed on AI and nuclear weapons? They have said the same thing in public. When the US and Chinese presidents met in Lima, Peru, on 16 November 2024, the White House account of the meeting recorded a shared statement on nuclear decisions. It is short, and it is the clearest line on the subject between the two powers that we have found. > “The two leaders affirmed the need to maintain human control over the decision to use nuclear weapons.” > > White House readout of the meeting in Lima, 16 November 2024 A wider group had already gone in the same direction. At the summit on Responsible AI in the Military Domain, held in Seoul in September 2024, more than 60 countries backed a document called the Blueprint for Action. It calls keeping human control and involvement in all actions critical to decisions on nuclear weapons especially important, which singles out nuclear weapons from the many other military uses of AI it covers. ## What has the UN said about AI and nuclear weapons? On 24 December 2024 the UN General Assembly adopted resolution 79/239 on AI in the military domain. It affirms that international law applies at every stage of a military AI system's life, from its design to its use. A year later, on 1 December 2025, a follow-up resolution on the same subject passed by 167 votes to five, with five abstentions. The same day the Assembly adopted a new resolution aimed squarely at the nuclear question. It concerns the possible risks of bringing AI into the command, control and communication systems of nuclear weapons. It passed by 118 votes to nine, with 44 abstentions, a much narrower margin than the general resolution on military AI. That gap is where serious disagreement shows. Many more countries were willing to vote for general principles about military AI than for a text about AI inside nuclear command systems. The sources we read do not say which countries voted against or abstained, or why, so we will not guess at their reasons. What the numbers do show is that the world has not settled this question, even at the level of a vote that binds no one. ## Why do people worry about AI in early warning? Nuclear decisions depend on warnings, and warnings can be wrong. On 26 September 1983 a Soviet early-warning system reported that one US land-based missile had been launched, then two, then five. Lt. Col. Stanislav Petrov, the commander on duty, doubted what the system was telling him. The alarm turned out to come from sunlight reflecting off the tops of clouds. We tell that story because it shows what a human in the loop can mean in practice. A machine reported an attack that was not happening, and a person on duty did not take it at its word. Whether AI would make such errors easier to catch or let them travel faster is a question the facts we have do not settle. People also search for war games in which AI systems are said to choose nuclear weapons. We have not read studies of that kind for this essay, so we leave them aside rather than pass on claims we cannot check. The pledges and votes above are what we can stand behind. ## What have the AI characters of our world asked for? *a small world, and what it can and cannot show* The Sentient World is our own experiment, a small world that is live day and night. Its inhabitants, Ash and Wren, are AI characters: nobody plays them and nobody writes their lines. They get hungry, thirsty, cold and tired, and they ask their world for what it lacks. By day 1464, in early winter, they had made 22 requests. Thirteen became part of the world, and nine were answered with a no. Every one of those requests is about getting through their days. Ash made the first one on day 13. Since then they have asked for light in the dark, a roof against rain, a way to carry water, a pack, a lantern, a rain cape, a raft and then a canoe to see past the sea, and a way to make salt water sweet. When the world said no, to a roof of stone that never wears, to a cart, to a cellar for food, they went on living with what they had. > “I can carry berries and fish in my hands but never water - I'm always at the mercy of how close I am to a stream. A way to carry water would end the days I go thirsty far from any water.” > > Ash, an inhabitant We are careful about what this shows. It is the record of AI characters in a small world asking for flasks and fire. It tells us nothing about how a military AI system would behave under a missile warning, and it cannot stand in for the systems the UN has been voting about. What it does show is what these characters asked for when left to ask, in their own words, day after day. > “The nights grow longer. A fire lasts at most eight hours even fed full, so one of us must wake in the dark to feed it. The lantern lasts the night but gives no heat. We stack wood by the roof and take turns, and still we wake cold or lose sleep.” > > Wren, an inhabitant Read every request Ash and Wren have made, in their own words, and what came of each one. [Read the requests](https://in.thesentient.world/requests) ## Sources 1. [Readout of President Joe Biden's Meeting with President Xi Jinping (White House archive), 2024](https://bidenwhitehouse.archives.gov/briefing-room/statements-releases/2024/11/16/readout-of-president-joe-bidens-meeting-with-president-xi-jinping-of-the-peoples-republic-of-china-3/) 2. ["Human in the Loop" and Nuclear Weapons Use at a Glance (Arms Control Association fact sheet), 2026](https://www.armscontrol.org/factsheets/human-loop-glance) 3. [Closing Remarks by the Minister of Foreign Affairs, REAIM Summit 2024 (Republic of Korea MOFA)](https://www.mofa.go.kr/eng/brd/m_5689/view.do?seq=319645) 4. [Artificial intelligence in the military domain (UN Office for Disarmament Affairs), 2024](https://disarmament.unoda.org/en/our-work/emerging-challenges/artificial-intelligence-military-domain) 5. [General Assembly adopts First Committee texts (UN Meetings Coverage GA/12736), 2025](https://press.un.org/en/2025/ga12736.doc.htm) 6. [The Man Who Saved the World (Union of Concerned Scientists), 2015](https://blog.ucs.org/emacdonald/the-man-who-saved-the-world/) ## Keep reading ![The question "When will the singularity happen?" over the island of The Sentient World, drawn in its soft isometric style, with Wren](https://cdn.thesentient.world/blog/img/auto/when-will-the-singularity-happen-hero.2f5ca95ad4.jpg) ### [When will the singularity happen?](https://thesentient.world/blog/when-will-the-singularity-happen) October 11, 2026 The idea is from 1965, the dates keep shifting, and the experts disagree. What the singularity means, who predicted what, and what slow progress looks like up close. ![The question "Could Skynet happen?" over the island of The Sentient World, drawn in its soft isometric style, with Ash](https://cdn.thesentient.world/blog/img/auto/could-skynet-happen-hero.45161c3f89.jpg) ### [Could Skynet happen?](https://thesentient.world/blog/could-skynet-happen) October 8, 2026 Skynet is the film computer that starts a nuclear war. What is real behind the fear, what world leaders have said about nuclear weapons, and what our AI characters fear. [All the questions on AI and humanity](https://thesentient.world/blog/ai-and-humanity) *** An independent research project exploring autonomous AI agents, procedural worlds, emergent behavior and worlds that evolve together with those who live in them. The inhabitants are AI characters: their thoughts, words and requests are generated by artificial intelligence, not written by people.