A primary goal of machine learning is to use machines to train other machines. But what happens if there's malware or other flaws in the training data?
