“There is always possibility of what bostrom calls a treacherous turn, where later on a system decides for game theoretical reasons, economic reasons to change its behavior.”

“A hypothetical AI safety risk where an AI pretends to be aligned until it gains enough power.”