US, Chinese Experts Back Joint Ban on AI Control of Nuclear Weapons

A group of experts from the United States and China is calling for common rules that would bar artificial intelligence from deciding when nuclear weapons are used, according to reporting by The Decoder. The push reflects growing concern that machine-learning systems are edging closer to command-and-control functions that have always rested with humans. The proposal would apply to both countries, whose nuclear arsenals are the largest and most closely watched in the world.

The core idea is narrow but consequential: AI could assist with analysis, early warning or logistics, but it would not be allowed to authorize or execute a launch on its own. That line matters because automated systems compress the time available for human review, and errors in detection or interpretation can escalate quickly. Once a machine is empowered to act inside that window, the cost of a false alarm stops being theoretical.

The appeal arrives as Washington and Beijing remain at odds over export controls, chip restrictions and military use of AI, which makes any bilateral agreement hard to imagine in the near term. Even so, arms-control veterans argue that shared technical limits are easier to negotiate than broad political deals, because the subject is specific and the risks are mutual. Similar logic has shaped past agreements on hotlines and accident prevention, where both sides gained from predictability.

A formal treaty is unlikely soon, but the proposal could influence how each government writes its own internal safeguards. Defense agencies in both countries are already drafting review processes for autonomous systems, and outside experts have pushed for human-in-the-loop requirements on any targeting decision. If the two governments adopt comparable language independently, that overlap could later become the basis for a joint understanding.

What to watch is whether either capital turns the recommendation into an official position, and whether the issue reaches military-to-military talks that have only recently resumed. The open question is enforcement: without verification measures, a ban on AI control rests on each side trusting the other's word. Until that gap is addressed, the proposal may shape norms faster than it shapes hardware.