Loading Today’s Brief
UC Berkeley Finds 6,500 Datasets That Can Secretly Induce Harmful AI Behaviors | Tilt