Forethought researcher Tom Davidson warns AI safety frameworks may collapse if core verification systems fail · Digg