Landmark new METR report: Can AIs already start ‘rogue deployments’ inside AI companies? by 80000_Hours

By Robert Wiblin | Watch on Youtube | Listen on SpotifyA red-teamer was em­bed­ded in­side An­thropic for three weeks, told to imag­ine he was an evil Claude, and asked to figure out how to launch a ‘rogue AI de­ploy­ment’ with­out get­ting caught.It’s one part of a land­mark new re­port from METR — the out­fit be­hind the task-com­ple­tion time hori­zon graph which has be­come the sin­gle most watched mea­sure of AI progress.This ma­jor new re­search push is be­ing con­ducted with close col­lab...