Ethics and safety·September 18, 2026, 11:37

US and China urged to ban AI control of nuclear weapons

AI-generated and checked against the sources listed below.

US and Chinese security experts are proposing joint rules to prevent artificial intelligence from making decisions about nuclear weapons. The proposal comes as a meeting between Trump and Xi is expected later in September.

AI-generated image

US and Chinese security experts have presented a set of joint rules meant to prevent artificial intelligence (computer programs that can analyze data and react to it on their own) from ever gaining control over nuclear weapons.

The proposal comes from Melanie Sisson of the US think tank Brookings Institution and Tianjiao Jiang of Fudan University in China. The two have written a report with concrete recommendations for how the US and China can set clear limits on how much power artificial intelligence may have in military and nuclear contexts.

Humans must stay in control

The core of the proposal is that no AI system should be able to decide on its own to launch nuclear weapons, and that AI also must not be used to attack the other side's systems for commanding and communicating about nuclear weapons. "Only humans, not AI, should make the decision to initiate a cyberattack against each other's nuclear command, control and communications systems," the authors write, according to the report.

The experts also propose that the two countries agree on a common definition of what "human control" actually means, and that a direct hotline be set up that can be used if something goes wrong with military AI. Concrete technical requirements include that a human must always be part of the decision chain, that equipment for launching nuclear weapons must be kept physically separate from networks, and that clear limits are set on how much machine learning may be used in command systems.

The background is that AI's speed could leave leaders with only a few minutes to react if a system began attacking the other side's nuclear weapons infrastructure on its own, something that according to the report could trigger a global catastrophe.

Trump and Xi expected to meet

The proposal builds on an agreement from November 2024, when then-President Joe Biden and Chinese President Xi Jinping agreed that humans, not AI, must have the final say when it comes to the use of nuclear weapons. The new report comes as a meeting between Donald Trump and Xi Jinping is expected to take place later in September, and the researchers hope the topic will be discussed there.

Neither government has officially endorsed the proposals yet, though. There is also reason for skepticism about whether the rules can really be enforced: Carla Freeman of Johns Hopkins University points out that China did not answer US calls during the so-called balloon crisis in 2023, which raises the question of whether a hotline between the two countries' militaries would even be used in a crisis.

For Danes, the issue is relevant because Denmark, as a NATO member, is closely tied to the great powers' handling of nuclear weapons and military AI. Clear international rules of the game can reduce the risk of errors or misunderstandings escalating into a nuclear conflict, something that ultimately also affects the security of Europe and Denmark.

Sources

More on this topic

Get the week's AI news in your inbox

Choose your level, topics and length. One email a week, unsubscribe at any time.

Subscribe to Promptly Newsletter
PromptlyNewsletterRSSLog in

The news on aijour is AI-generated and checked against the cited sources.