Algorithmic Fairness and ML Safety