Either that, or it is happy that it was correct and not merely overestimating him resulting in a disproportionate response. Because that would mean the AI failed utterly in minimizing the cost function f(x) [with the constraint g(x)].
And none likes failing at their core directives~