The idea that a hostile Super-AI will take over and/or be a Doomsday scenarios is equally plausible the reverse will be equally irreversible.

A benevolent AI would immediately prioritize:

A) Eliminating hostile AI research

B) Advancing the good of humanity on the metrics it was given.