The Most Dangerous Bias of Your AI Assistant Is That It Agrees with You – Part 2: Why We Also Need to Remove Rules Again