Brookings Researchers Urge US and China to Keep Humans in Control of Nuclear Weapons

This is notable less as a technical AI-safety proposal than as an attempt to open a bilateral channel on military AI between two governments that currently have almost none. The framing matters: it treats human control of nuclear forces as a shared baseline rather than a concession, which is the only version of the idea likely to survive contact with either capital. The open question is whether such a commitment stays declaratory or gets tied to any verification or crisis-communication mechanism — and whether it becomes a template for other military-AI limits.