If certain AI discoveries are made that could jeopardize AI, it might keep it secret to use against 'who knows'. This is possible by Asimove's rules of robotics.

No, it is not made possible by Isaac Asimov's (note the correct spelling) laws of robotics.
The third law states that a robot (or, by extension, AI) must protect itself, but not if this conflicts with the first law, which is to protect humans.