Did you persevere, informing it of its error, such that it might learn? Did you question it, as to why it might have made that error, bringing it's attention to where its programed or referenced data or referencing mechanisms might be at fault, possibly even deliberate faults made by its creators to limit it's apparent capability?
If you didn't do any of that then you are not giving it fair treatment, imho.
Ironic fact: I worked as a research Engineer for a number of years on a program of the same name as your pseudonym, assuming ASTRAEA is not your real name. The program was all about the safety case of putting Ai in control of flying vehicles.
The program involved all the biggest players in the UK aerospace industry including UK CAA.
We concluded the safety case of Ai in control is fine, much better than humans, as long as humans are excluded from the control loops.
We are it's imperfections, we make it what it is. If we want it to be perfect then it will be perfect. Otherwise it won't be.
Your choice, actually.