People will not be held responsible; the disasters caused by AI will be attributed as a natural phenomenon -- like the weather. The companies involved in creating and operating AI will certainly not take any accountability for any "spills".
I don't know how you can confidently say that without first knowing who the victims are. If AI causes a disaster that harms a member of the American billionaire class, someone will need to satisfy their bloodlust.
I'm excited for AI executives declaring private military action against each other.
OpenAI and Antropic are trying to create that reality, but it would be ideal if they failed.