I am wondering does anyone have a survey or list of AI exploits or malicious actions done on production services or systems ? For example like if a misclassified image that would target a image recognition system (such as Clarfai)? I have only seen papers that have theoretical attacks so far.