Because there is so much short-term incentive to deploy AI (for business and military, esp) there may be a tendency to rush into it without consideration of the down side (assessing the risks). 

It's become a similar situation to the development of nukes. The risk assessments were done with a keen focus of the down-side of NOT deploying. The risk of NOT using it was deemed greater than the risk of using it. There is a similar situation here. 

This also recalls Carl Sagan's famous "shiny red button" hypothesis. He speculated that the solution to the Fermi Paradox (if there are superior non-human races out in the cosmos, why haven't we seen them yet?) is that any technological civilization may discover some technology that is irresistible (like a shiny red button at a science exhibit) and when pressed causes catastrophic destruction of the culture using it. BOOM!!

That view is technological development that is sufficiently sophisticated always results in the destruction of such cultures before they can travel to the stars.