Assuming the plane is on autopilot: for example, if a plane is flying high (over 20000 feet) and its autopilot receives a positive value of external temperature, which is clearly abnormal or impossible.
What will the autopilot do with such a obviously wrong information?
The autopilot does not receive information about the outside temperature, it has no (direct) relation to what the autopilot controls.
The autopilot receives aircraft state inputs like airspeed, Mach number, vertical speed, attitude, accelerations etc.
I can not tell you what happens if there is a positive value for the temperature above 20000ft, but here is some general information:
For this issue I will be talking about Airbus planes, as they are known to be very "smart" planes. There are hundreds of algorithms that try to prevent wrong data input from being an issue. But those do not always work.
Take Qantas Flight 72 for example: According to Wikipedia the airplane all of the sudden pitched down violently, as if it was trying to prevent a stall from happening. This was because the average of the two angle of attack sensors was way out the box. Though there was a (very smart) algorithm rechecking the values, it could not deal with multiple wrong values with a certain amount of time - and made the airplane pitch down.
This being an example where the plane actually acted by itself due to a faulty value, here is the general case: If the airplane notices odd values, which it will usually recognize (e.g. a positive temperature), it blanks out the display so that a pilot can not be mislead.
When flying smaller airplanes, you often practice "partial panel flying", where your flight instructor covers up some instruments and you have to fly the aircraft with alternative instruments. As you can see this is well practiced and shouldn't be a problem. Unfortunately though, this makes flying a lot harder which can cause crashes to happen (AF447,QZ8501, and more).
Furthermore, whenever the autopilot finds itself in a situation that it can't handle, for example when the pitot static system fails, it will disconnect and refuse to connect, giving the pilot full control (or rather as full control as possible on an Airbus) of the aircraft. This disengagement ensures the plane not to do something stupid due to wrong inputs.
Conclusion: For redundancy most systems exist two or three times working parallel to provide backup, same goes for sensors. If all data inputs report something obviously wrong, the pilot will not see this wrong information so he can concentrate on what he has.
The answers to this question can also give you a little more information about computer redundancy on aircraft
Depending on the 'architecture' or system design of the avionics, the autopilot may or, at today's levels of technology, may not directly receive temperature information.
The Temperature is taken from multiple systems of TAT (Total Air Temperature) probes and is fed to an ADC (Air Data Computer) or equivalent, which then analyses the inputs received for bad data and it will typically have redundancies that will allow it to discard the bad and provide only good data to it's 'clients', one client being the autoflight system.
Depending on the architecture and system features, the autoflight system or subsystem may even receive both, i.e. from the ADC and direct inputs from a local probe.
With the advent of FBW, Fly-by-wire, all data comes in for scrutiny by some computer or the other, and probably more than once.
Whether bad data can get past all the redundancies and cause an autopilot to misbehave, we can only answer that the chances are remote, but it can happen. Regarding a temperature input that is +ve above 20,000ft - this may not always be way out of line as, depending on the speed, the TAT could conceivably be around the 0degC mark at those levels at around Mach 0.7
too high (about 20 tousand feets)? What do you mean byenter in a new procedure or ask itself that it is not normal? $\endgroup$