In William Gibson's prophetic 1984 book Neuromancer, he described AI that could autonomously wage cyber warfare. Today, we are regularly warned about the security and privacy implications of software that can learn and make decisions from massive datasets, but rarely do we discuss the possibility of this same software being used to make cyber attacks, or the possibility that it could escape our control. Through a series of questionable ethical decisions, the first iteration of just that has been developed. WinterMute is an open-source deep learning project that can assess and exploit a network, reproduce itself, pivot on a compromised machine, and learn from each step.