In what seems like HAL 9000 come to malevolent life, a recent study appeared to demonstrate that AI is perfectly willing to indulge in blackmail, or worse, as much as 89% of the time if it doesn’t get its way or thinks it’s being switched off. Or does it?