In the movie "Ex Machina" we reveal that Nathan has in fact created multiple AI robots before Ava, many which had started to express hostility towards him. We would think that when creating Ava, he would include a "master override password" (not even getting into the 3 Laws of Robotics) of some sorts that when spoken, would completely disable the AI. It seems very unusual that Nathan would either overlook or intentionally exclude this safety percaution. Was this because he wanted to make Ava as realistic as possible or for some other reason?
It's been a while since I watched Ex Machina, but I remember Nathan's arrogance. He thought he was in control of everything. He thought he was several steps ahead of his 'bots and his human guest. He thought he had them trapped in the environment he had created. He also thought that since he had created his 'bots, he was their "master".
To his narcissistic mind, there was no need to put in a fail-safe like a master override password. (I don't have any outside sources to back up my conclusion, but it fits with Nathan's personality.)
This is something even I had thought of. Nathan being the genius he is, could have just programmed an automatic switch off if the robot crossed a boundary line or he could have even had a wireless switch off mechanism. Given that another robot had broken a glass pane and was desperate to exit the building, Nathan should have indeed built in a safety feature. However I don't entirely agree with BrettFromLA's perspective that it was because of Nathan's arrogance. Nathan did take few basic precautions. The access cards, the underground facility with no windows, the fact that he did a lot of weight training that could enable him to overpower the robots easily.
I would assume it'd have to do with the way he built the robots. If somebody built a kill switch onto your body, you'd eventually find out. The robot's brains were built in a way that's similar to the human brain. If any of those highly intelligent circuits detected any unknown mechanisms, the robot would strive to override it in the same way Ava learnt to perform power shutdowns. Also, the objective was to allow Ava to utilize everything she could, to escape. The previous robots didn't really achieve the purpose and Nathan may have not really imagined that Ava was yet capable of escaping. You can see that in his conversations with Caleb, where he keeps asking him of what he thought of the AI.
But these are just guesses. It's the filmmaker who should answer this question. Maybe it was just a deliberate oversight for the sake of creating a chilling climax that gripped the audience.