What happens when a robot does something that it isn't programmed to do?

[DSMOS]

I just finished watching Big Hero 6, and a part that really caught me was when Hiro removed Baymax's original programming, but when it was put back in, he apologized. Then when Hiro tried to remove it again, Baymax wouldn't let him. How does this work? It feels like one of those "I'm afraid I cannot let you do that, Dave" moments...

[mklrobo]

  Hello!
A complex question, to be sure.
The answer lies in the complexity of the robot. The more "layers"
of programming and MCUs, the more systems that can inhibit
an action.
In the case of the scenario you indicated, I would think that Baymax had
"learned" what had happened, and did not want to be rebooted. Back in the day,
there was ROM, (Read Only Memory) which could not be written over.
I would hope, in the fiction or fact world, that the main "overseeing"
AI that controls the robot is not damaged, and maintains a divine
integrity.(assuming the robot is good, of course. NOT Ultron!)

[mklrobo]

  Hello!
Just saw the movie, and now I can fully answer your question;
and a part that really caught me was when Hiro removed Baymax's original programming, but when it was put back in, he apologized. Then when Hiro tried to remove it again, Baymax wouldn't let him.
In context, there are 2 parts in the movie; when they were in a fight scene, nobody could get to the data port
until the battle ended}?{ @ the end, when they were in hyperspace, the original baymax stayed behind,
but his "mind" was in the card.(programming) In my opinion, baymax did not wieght anything, so why not
let him go along through the portal? Alot of weak moments written to generate deep sorrow, which works
if the story is written properly, but in this case, leaves the audience with more questions than closure.}(?)