Experts from the United States and China have published proposals calling for shared rules that would prohibit AI systems from making autonomous decisions about deploying nuclear weapons. The recommendations, from Melanie Sisson of the Brookings Institution and Tianjiao Jiang of Fudan University, arrived ahead of a planned meeting between President Donald Trump and President Xi Jinping on September 24.
Red Lines for Nuclear Command
The proposals spell out several red lines. No AI system should be allowed to launch nuclear weapons on its own. No AI system should be allowed to attack nuclear command systems. Humans must retain sole control over AI-driven cyberattacks on strategic infrastructure. Both countries should also agree on a shared definition of "human control," a term the two sides currently interpret differently.
The recommendations build on an agreement Biden and Xi reached in November 2024. That earlier understanding set the stage for the current push, but it left the hardest questions unresolved.
"Binding mechanisms still don't exist," the authors note, pointing to a gap between stated intentions and enforceable limits.
A Hotline for AI Incidents
Jiang also proposes a hotline for AI incidents, modeled on Cold War-era direct lines between Washington and Moscow. The logic is straightforward. If a defensive AI system automatically responds to suspicious activity, the other side might read it as an attack. A direct line would let one government clarify that an incident was accidental before the situation escalates.
The idea is not new. The National Security Commission on Artificial Intelligence (NSCAI) called for preserving human control over nuclear weapons back in 2021. Nearly five years later, binding mechanisms still don't exist, and the commission's recommendations remain unfulfilled.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Skepticism About Crisis Communication
Carla Freeman of Johns Hopkins University questions whether such hotlines would even work. During the spy balloon crisis in 2023, China didn't pick up when the US called. A direct line, she suggests, is only useful if someone answers.
The concern cuts to the heart of the proposals. Shared rules require shared trust, and the historical record offers little comfort. The 2023 episode showed how quickly a single incident can strain communication channels that both sides claim to value.
What Comes Next
The Trump-Xi meeting on September 24 will test whether the two governments can move from principles to practice. The proposals from Sisson and Jiang offer a concrete starting point, but they remain recommendations rather than commitments. Without binding mechanisms, the red lines they describe depend on political will that has yet to materialize.
For now, the experts' work stands as a blueprint. Whether it becomes policy depends on decisions made in Washington and Beijing, far from the think tanks and university offices where the proposals were written.