But this time, focused on AI safety (just like all the previous ones).