I think that there can be strategies that make AI the more safer the more capable it becomes. There was a time when I became safer when I became more capable as I thought, reflected and kept certain actions off limits. There can be core and basic stratgies to be used for making AI safer the more capable it becomes and now we are going in the opposite direction with about 50% probabality. 

There must be some humans that atleast for a point of time became safer the more capable they became and their strategies whichever are compatiable for AI can be used. Instead of all being black and white there must be certain tyoes of capablities that make AI safer instead of more dangerous. Editing the concept of AI to make it no longer be AI that is better than AI could help as AI might not be the only threat that people could face as in future there could be more dangerous concepts created like people could create variations of the concept of AI that are more dangerous for their time. 

Instead of banning AI the concept of AI if it becomes obselete and safer with another concept could help. People might need stratgies to help better than banning AI ever can. 

There could be atleast 3 types of concepts that are more dangerous AI successor, less dangerous AI successor and equally dangerous AI successor and if a concept is made that is less dangerous AI successor that makes AI obselete that may help although I an not sure if there is moderate work in making AI obselete. 

Humans had handled harmful substances that together make useful substances and if that concept could be fused with AI making it different but safe like an safety booster then that could help and that safetly booster only changed the concept of AI after being fused with it then it might be progess and a strategy could be instead of having to contain AI that is smart it could be combined with another concept perhaps by adding additional lines of code to it to make it no longer AI and change the problem of contaning AI to contaning the Fused AI concept and if that is not smart that "that could be better" and it might not need to be smart. 

A strategy of contaning AI might be to no longer make it be an AI by adding additional lines of code to it but that might fail if those lines of code are removed. However it could be another strategy and better than nothing and it might be like instead of trying to contain the spread of poisionous gass in a place by releasing gas that merges with that poisions gas and makes it harmless and it might had been done before with precedent. This is not medical advice.

I guess that there might be only a few ways that superintelligence AI could harm us and if those ways are stopped then it might help but it could be like hypothetically trying to prevent a slightly posionous snake from trying to harm a man and it might be better to add a guard to snake's mouth to prevent it from biting and spitting poison and snake catchers have existed with snakes for a long time perhaps but I hardly if ever had heard the news of a snake catcher being bitten by a snake. I have heard more cases of snakes biting ordinary people and snake catchers might be using strategies to be avoid being bitten by snakes given their number(of the snake catchers and the number of snake catchers perhaps an estimate was probabaly written in an old NCERT of class 7-10(In Social Studies NCERT if I am right). 

The topic of this quick take is AI superintelligence and avoidance strategies.