Seems to me that the goal is to build a funding model. There CANNOT be such a thing as "Safe Superintelligence". A ML system can ALWAYS (by definition of ML) be exploited to do things which are detrimental to consumers.