Exploring whether AI can hide its true motives and the implications for safety and transparency in machine learning.