Adversarial machine learning

Adding malicious inputs to an AI algorithm in order to intentionally trick it.