If they simulate a massive enough neural net, fast enough with real time inputs/outputs, on a long enough time-frame, what to do if it becomes self-conscious?

If they simulate a massive enough neural net, fast enough with real time inputs/outputs, on a long enough time-frame, what to do if it becomes self-conscious? Pulling the plug would be brutal murder. That said, it seems it is nearly impossible, even for a supercomputer to get close to the "size"/"speed" of one human brain which seems to be the basic requirements for consciousness to emerge.