Executives at leading artificial intelligence firms are privately gaming out worst-case scenarios and public revolt strategies ahead of a potential catastrophic AI incident.