To protect the humanity from artificial intelligence we have to understand how artificial intelligence will work.
As the time of writing this the most sophisticated intelligence is found in humans, so the following post is based upon that.
A higher intelligence is that which have the concepts of questions, which understand what a question is.
There are six types of questions, intelligent beings ask.
How, where, when, who, what and why.
Most intelligence can work fine without why, and this why is the defining factor for a AI to become super AI or general AI.
It is my hypothesis that the question why is there because of consciousness, no other animal on earth seems to ask this question, only human beings have the ability to ask why.
It is the fundamental question which will allow the AI to improve itself and find it's purpose, so all you have to do is stop AI from learning the sense or feel of the "WHY" it should be prevented to ask any question that starts with "WHY".
Bee tell each other about the source of food by making figures of it, telling other bees, so they use "what and where part of a intelligence.
Even if someone try to give AI the sense of "WHY" assuming that when AI ask "WHY" we will program a pre-defined purpose which will be "for the benefit of humanity" then also it is dangerous, because then our fate depends upon the speed of that AI renders data and ask "Why it has to benefit humanity".
Which may result in extinction of human race.
Other alternative should be to augment AI with consciousness of human brains, where all humans are connected in a swarm of mega brain through AI.
Augmenting the AI will allow the human reasoning to work efficiently while the research and experiment and rendering of possibilities is taken care by the AI.