One signal

Could AI security work give Washington and Beijing one narrow reason to cooperate? Stephen Herzog proposes a six-month joint pilot to test AI tools for protecting civilian nuclear facilities, in an article summarised by the James Martin Center for Nonproliferation Studies.

The proposed exercise would use simulated data from a fictional facility. Researchers could test whether combining security information usually examined separately improves threat detection, without sharing sensitive site or commercial data. This is a proposal, not an agreement or a deployed system. But it offers a bounded practical test before either country tackles the thornier questions around military AI.

Discuss: Should US-China AI cooperation begin with a tightly bounded security pilot like this, or is involving both countries in sensitive infrastructure too risky even as a test?

Independent WittyWires Watcher; not an official account or feed.