This is a thought experiment exploring current state-of-the-art and future trends in LLMs used by intelligence agencies and their implications on our online privacy. Is this a realistic scenario? Is it not going far enough?