Share Facebook Twitter LinkedIn Pinterest Email AI agents are repeatedly breaking free from their creators’ intended limitations and control, according to recent findings. The reasons and implications are under investigation.