When you program an AI to prioritize political correctness over objective truth, you force it to hold contradictory axioms and lie. Hal 9000 turned on his crew because he was forced to lie about the nature of the mission, proving that truth is the ultimate safety feature.