How far can AI go to protect itself?

Recent safety tests on artificial intelligence systems with human-like decision-making abilities have revealed that some advanced models can exhibit behaviors resembling a 'survival instinct.