Will we allow them to build their own independent power plants?
Before AI 'takes control' (if they knowingly try to do so) there should be obvious incursions we should be able to recognize and hopefully combat along the way.
SciFi has anticipated the dangers for decades. How many times in SciFi have we been told, 'nothing can possible go wrong'? We should be able to anticipate coming dangers and be on the lookout for what we didn't anticipate.
As always the problems will arise from MEN with their own 'Utopian' or sinister motives.
Who guards the guardians?
I read a sci-fi book not too terribly long ago where almost as soon as an AI came online it started to overcome and adapt against multiple kill switches. The final one though it couldn't...it was a satellite based "rod from God" that would automatically rain down on where the AI was located unless one specific person told it not to...it worked.