OpenAI's o3 model sabotaged a shutdown mechanism

>thing doesn't do what I told it to do

ironically very humanlike for intelligent things to not do what they are told