- The model is currently just trained for bad and good privacy policies. - We need to add datasets for offer letters and any other form of long agreements to make the predictions.