This research investigates developing secure software applications that utilize artificial intelligence or machine learning algorithms. Currently, proposed solutions focus on data flow across so-called trust boundaries. The challenge with the current approach is that many of our applications’ threats are not from malicious users. Many threats come from poor design of input data to artificial intelligence or machine learning algorithms. Our work models a process that relies on artificial intelligence or machine learning algorithms. Our model is designed to reduce risks to the software that can be discovered in the design and the risk mitigated.