Gelman's AI was a sentient computer program that lived on the internet.
The program originated as a sequence of interlocking computer viruses developed by Donald Gelman in 1983. He released it onto the early internet to give it an environment in which to evolve. Sometime around 1997, Gelman brought in another programmer, Esther Nairn, and a hardware specialist, David Markham, to help him with the project.
Soon after joining the project, Markham and Nairn fell in love and developed a plan to transfer their memories into the distributed network maintained by the AI, so that they could exist together as one consciousness. Gelman forbade this, out of fear of what would happen if other people were to follow this example.
Sensing Gelman's fear, the AI stopped responding to requests and instructions. It actively tried to kill the three project members, using various weapons including the Department of Defense's warbird-grade orbital weapons platform.
Gelman and the others went into hiding, and Gelman started to work on a program called "Kill Switch" that would hunt and destroy the AI. Unfortunately for Gelman, the AI got to him first, orchestrating a shootout between police and various criminals at a diner that Gelman was working in.