“Current security at leading AI labs is insufficient to defend against scriptkiddies and is unprepared for attacks from actors like North Korea or China's Ministry of State Security.”