Comment by AntiDyatlov

A very good outcome for AI safety would be if when improved models get released, malicious actors use them to break society in very visible ways. Looks like we're getting close to that world.