He said AI algorithms learn and improve themselves through identifying pattern in training data and the entire algorithm would be vulnerable to similar bias. "Another risk of AI implementation in the public sector is privacy concerns for the individuals."