Bessent told reporters after the talks that Washington had proposed a new AI safety notification mechanism and a new US-China AI dialogue, with a particular emphasis on national security concerns.